IBM Certification Study Guide
AIX HACMP
David Thiessen, Achim Rehor, Reinhard Zettler
International Technical Support Organization
http://www.redbooks.ibm.com
SG24-5131-00
Download from Www.Somanuals.com. All Manuals Search And Download.
SG24-5131-00
International Technical Support Organization
IBM Certification Study Guide
AIX HACMP
May 1999
Download from Www.Somanuals.com. All Manuals Search And Download.
Take Note!
Before using this information and the product it supports, be sure to read the general information in
First Edition (May 1999)
This edition applies to HACMP for AIX and HACMP/Enhanced Scalability (HACMP/ES), Program
Number 5765-D28, for use with the AIX Operating System Version 4.3.2 and later.
Comments may be addressed to:
IBM Corporation, International Technical Support Organization
Dept. JN9B Building 003 Internal Zip 2834
11400 Burnet Road
Austin, Texas 78758-3493
When you send information to IBM, you grant IBM a non-exclusive right to use or distribute the
information in any way it believes appropriate without incurring any obligation to you.
© Copyright International Business Machines Corporation 1999. All rights reserved.
Note to U.S Government Users – Documentation related to restricted rights – Use, duplication or disclosure is
subject to restrictions set forth in GSA ADP Schedule Contract with IBM Corp.
Download from Www.Somanuals.com. All Manuals Search And Download.
Contents
Figures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ix
Tables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xi
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xiii
2.1 Cluster Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Cluster Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Cluster Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.1 SSA Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.2 SCSI Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4 Resource Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.5 Application Planning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.7 User ID Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
© Copyright IBM Corp. 1999
iii
Download from Www.Somanuals.com. All Manuals Search And Download.
3.1 Cluster Node Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.3 Cluster Disk Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.1 SSA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.3.2 SCSI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.5 Quorum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.1 Installing HACMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.1.1 First Time Installs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.3 Defining Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.4 Initial Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.4.1 Clverify . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.4.2 Initial Startup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.5 Cluster Snapshot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
iv
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
5.2 Error Notification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
5.4 NFS considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
5.4.5 Cross Mounted NFS File Systems and the Network Lock Manager.
128
6.1 Node Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.2 Simulate Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.2.4 Disk Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
7.1 Cluster Log Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
7.2 config_too_long . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
7.3 Deadman Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
7.6 User ID Problems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.1 Monitoring the Cluster . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
v
Download from Www.Somanuals.com. All Manuals Search And Download.
8.3.1 Nodes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.3.2 Adapters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
8.3.3 Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.4.3 C-SPOC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
8.7 Backup Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
9.2 Kerberos Security. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
vi
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
9.3 VSDs - RVSDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
10.5 Decision Criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
B.2 Redbooks on CD-ROMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
B.3 Other Publications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
List of Abbreviations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
vii
Download from Www.Somanuals.com. All Manuals Search And Download.
viii
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Figures
1. Basic SSA Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2. Hot-Standby Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3. Mutual Takeover Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5. Single-Network Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6. Dual-Network Setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7. A Point-to-Point Connection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
9. Connecting Networks to a Hub . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
15. Applying a PTF to a Cluster Node . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
© Copyright IBM Corp. 1999
ix
Download from Www.Somanuals.com. All Manuals Search And Download.
x
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Tables
8. SSA Disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
9. SSA Adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
21. HACMP Log Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
© Copyright IBM Corp. 1999
xi
Download from Www.Somanuals.com. All Manuals Search And Download.
xii
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Preface
The AIX and RS/6000 Certifications offered through the Professional
Certification Program from IBM are designed to validate the skills required of
technical professionals who work in the powerful and often complex
environments of AIX and RS/6000. A complete set of professional
certifications is available. It includes:
• IBM Certified AIX User
• IBM Certified Specialist - RS/6000 Solution Sales
• IBM Certified Specialist - AIX V4.3 System Administration
• IBM Certified Specialist - AIX V4.3 System Support
• IBM Certified Specialist - RS/6000 SP
• IBM Certified Specialist - AIX HACMP
• IBM Certified Specialist - Domino for RS/6000
• IBM Certified Specialist - Web Server for RS/6000
• IBM Certified Specialist - Business Intelligence for RS/6000
• IBM Certified Advanced Technical Expert - RS/6000 AIX
Each certification is developed by following a thorough and rigorous process
to ensure the exam is applicable to the job role and is a meaningful and
appropriate assessment of skill. Subject Matter Experts who successfully
perform the job participate throughout the entire development process. These
job incumbents bring a wealth of experience into the development process,
thus making the exams much more meaningful than the typical test, which
only captures classroom knowledge. These Subject Matter experts ensure
the exams are relevant to the real world and that the test content is both
useful and valid. The result is a certification of value that appropriately
measures the skill required to perform the job role.
This redbook is designed as a study guide for professionals wishing to
prepare for the certification exam to achieve IBM Certified Specialist - AIX
HACMP.
The AIX HACMP certification validates the skills required to successfully
plan, install, configure, and support an AIX HACMP cluster installation. The
requirements for this include a working knowledge of the following:
• Hardware options supported for use in a cluster, along with the
considerations that affect the choices made
xiii
Download from Www.Somanuals.com. All Manuals Search And Download.
• AIX parameters that are affected by an HACMP installation, and their
correct settings
• The cluster and resource configuration process, including how to choose
the best resource configuration for a customer requirement
• Customization of the standard HACMP facilities to satisfy special
customer requirements
• Diagnosis and troubleshooting knowledge and skills
This redbook helps AIX professionals seeking a comprehensive and
task-oriented guide for developing the knowledge and skills required for the
certification. It is designed to provide a combination of theory and practical
experience. It also provides sample questions that will help in the evaluation
of personal progress and provide familiarity with the types of questions that
will be encountered in the exam.
This redbook will not replace the practical experience you should have, but,
when combined with educational activities and experience, should prove to
be a very useful preparation guide for the exam. Due to the practical nature of
the certification content, this publication can also be used as a desk-side
reference. So, whether you are planning to take the AIX HACMP certification
exam, or just want to validate your HACMP skills, this book is for you.
For additional information about certification and instructions on How to
Register for an exam, call IBM at 1-800-426-8322 or visit our Web site at:
http://www.ibm.com/certify
The Team That Wrote This Redbook
This redbook was produced by a team of specialists from around the world
working at the International Technical Support Organization Austin Center.
David Thiessen is an Advisory Software Engineer at the International
Technical Support Organization, Austin Center. He writes extensively and
teaches IBM classes worldwide on all areas of high availability and clustering.
Before joining the ITSO six years ago, David worked in Vancouver, Canada
as an AIX Systems Engineer.
Achim Rehor is a Software Service Specialist in Mainz/Germany. He is Team
Leader of the HACMP/SP Software Support Group in the European Central
Region (Germany, Austria and Switzerland). Achim started working with AIX
in 1990, just as AIX Version 3 and the RISC System/6000 were first being
introduced. Since 1993, he has specialized in the 9076 RS/6000 Scalable
xiv
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
POWERparallel Systems area, known as the SP1 at that time. In 1997 he
began working on HACMP as the Service Groups for HACMP and RS/6000
SP merged into one. He holds a diploma in Computer Science from the
University of Frankfurt in Germany. This is his first redbook.
Reinhard Zettler is an AIX Software Engineer in Munich, Germany. He has
two years of experience working with AIX and HACMP. He has worked at IBM
for two years. He holds a degree in Telecommunication Technology. This is
his first redbook.
Thanks to the following people for their invaluable contributions to this
project:
Marcus Brewer
International Technical Support Organization, Austin Center
Rebecca Gonzalez
IBM AIX Certification Project Manager, Austin
Milos Radosavljevic
International Technical Support Organization, Austin Center
Comments Welcome
Your comments are important to us!
We want our redbooks to be as helpful as possible. Please send us your
comments about this or other redbooks in one of the following ways:
to the fax number shown on the form.
• Use the electronic evaluation form found on the Redbooks Web sites:
For Internet users
For IBM Intranet users
http://www.redbooks.ibm.com
http://w3.itso.ibm.com
• Send us a note at the following address:
xv
Download from Www.Somanuals.com. All Manuals Search And Download.
xvi
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 1. Certification Overview
This chapter provides an overview of the skill requirements for obtaining an
IBM Certified Specialist - AIX HACMP certification. The following chapters
are designed to provide a comprehensive review of specific topics that are
essential for obtaining the certification.
1.1 IBM Certified Specialist - AIX HACMP
This certification demonstrates a proficiency in the implementation skills
required to plan, install, and configure AIX High Availability Cluster
Multi-Processing (HACMP) systems, and to perform the diagnostic activities
needed to support Highly Available Clusters.
Certification Requirement (two Tests):
To attain the IBM Certified Specialist - AIX HACMP certification, candidates
must first obtain the AIX System Administration or the AIX System Support
certification. In order to obtain one of these prerequisite certifications, the
candidate must pass one of the following two exams:
Test 181: AIX V4.3 System Administration
or
Test 189: AIX V4.3 System Support.
Following this, the candidate must pass the following exam:
Test 167: HACMP for AIX V 4.2.
Recommended Prerequisites
A minimum of six to twelve months implementation experience installing,
configuring, and testing/supporting HACMP for AIX.
Registration for the Certification Exam
For information about how to register for the certification exam, please visit
the following Web site:
http://www.ibm.com/certify
© Copyright IBM Corp. 1999
1
Download from Www.Somanuals.com. All Manuals Search And Download.
1.2 Certification Exam Objectives
The following objectives were used as a basis for what is required when the
certification exam was developed. Some of these topics have been
regrouped to provide better organization when discussed in this publication.
Section 1 - Preinstallation
The following items should be considered as part of the preinstallation plan:
• Conduct a Planning Session.
• Set customer expectations at the beginning of the planning session.
• Gather customer's availability requirements.
• Articulate trade-offs of different HA configurations.
• Assist customers in identifying HA applications.
• Evaluate the Customer Environment and Tailorable Components.
• Evaluate the configuration and identify Single Points of Failure (SPOF).
• Define and analyze NFS requirements.
• Identify components affecting HACMP.
• Identify HACMP event logic customizations.
• Plan for Installation.
• Develop a disk management modification plan.
• Understand issues regarding single adapter solutions.
• Produce a Test Plan.
Section 2 - HACMP Implementation
The following items should be considered for proper implementation:
• Configure HACMP Solutions.
• Install HACMP Code.
• Configure an IP Address Takeover (IPAT).
• Configure non-IP heartbeat paths.
• Configure a network adapter.
• Customize/tailor AIX.
• Set up a shared disk (SSA).
• Set up a shared disk (SCSI).
• Verify a cluster configuration.
2
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Create an application server.
• Set up Event Notification.
• Set up event notification and pre/post event scripts.
• Set up error notification.
• Post Configuration Activities.
• Configure a client notification and ARP update.
• Implement a test plan.
• Create a snapshot.
• Create a customization document.
• Perform Testing and Troubleshooting.
• Troubleshoot a failed IPAT failover.
• Troubleshoot failed shared volume groups.
• Troubleshoot a failed network configuration.
• Troubleshoot failed shared disk tests.
• Troubleshoot a failed application.
• Troubleshoot failed Pre/Post event scripts.
• Troubleshoot failed error notifications.
• Troubleshoot errors reported by cluster verification.
Section 3 - System Management
The following items should be considered for System Management:
• Communicate with the Customer.
• Conduct a turnover session.
• Provide hands-on customer education.
• Set customer expectations of their HACMP solution's capabilities.
• Perform Systems Maintenance.
• Perform HACMP maintenance tasks (PTFs, adding products, replacing
disks, adapters).
• Perform AIX maintenance tasks.
• Dynamically update the cluster configuration.
• Perform testing and troubleshooting as a result of changes.
Certification Overview
3
Download from Www.Somanuals.com. All Manuals Search And Download.
1.3 Certification Education Courses
Courses and publications are offered to help you prepare for the certification
tests. These courses are recommended, but not required, before taking a
certification test. At the printing of this guide, the following courses are
available. For a current list, please visit the following Web site:
http://www.ibm.com/certify
Table 1. AIX Version 4 HACMP Installation and Implementation
Course Number
Course Duration Five days
Course Abstract This course provides a detailed understanding of the
Q1054 (USA) AU54 (Worldwide)
High Availability Clustered Multi-Processing for AIX.
The course is supplemented with a series of laboratory
exercises to configure the hardware and software
environments for HACMP. Additionally, the labs provide
the opportunity to:
• Install the product.
• Define networks.
• Create file systems.
• Complete several modes of HACMP installations.
4
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The following table outlines information about the next course.
Table 2. AIX Version 4 HACMP System Administration
Q1150 (USA); AU50 (Worldwide)
Five days
Course Number
Course Duration
Course Abstract
This course teaches the student the skills required to
administer an HACMP cluster on an ongoing basis
after it is installed. The skills that are developed in this
course include:
• Integrating the cluster with existing network
services (DNS, NIS, etc.)
• Monitoring tools for the cluster, including HAView
for Netview
• Maintaining user IDs and passwords across the
cluster
• Recovering from script failures
• Making configuration or resource changes in the
cluster
• Repairing failed hardware
• Maintaining required cluster documentation
The course involves a significant number of hands-on
exercises to reinforce the concepts. Students are
expected to have completed the course AU54
(Q1054) HACMP Installation and Implementation
before attending this course.
Certification Overview
5
Download from Www.Somanuals.com. All Manuals Search And Download.
6
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 2. Cluster Planning
The area of cluster planning is a large one. Not only does it include planning
for the types of hardware (CPUs, networks, disks) to be used in the cluster,
but it also includes other aspects. These include resource planning, that is,
planning the desired behavior of the cluster in failure situations. Resource
planning must take into account application loads and characteristics, as well
as priorities. This chapter will cover all of these areas, as well as planning for
event customizations and user id planning issues.
2.1 Cluster Nodes
One of HACMP’s key design strengths is its ability to provide support across
the entire range of RISC System/6000 products. Because of this built-in
flexibility and the facility to mix and match RISC System/6000 products, the
effort required to design a highly available cluster is significantly reduced.
In this chapter, we shall outline the various hardware options supported by
HACMP for AIX and HACMP/ES. We realize that the rapid pace of change in
products will almost certainly render any snapshot of the options out of date
by the time it is published. This is true of almost all technical writing, though
to yield to the spoils of obsolescence would probably mean nothing would
ever make it to the printing press.
The following sections will deal with the various:
• CPU Options
• Cluster Node Considerations
available to you when you are planning your HACMP cluster.
2.1.1 CPU Options
HACMP is designed to execute with RISC System/6000 uniprocessors,
Symmetric Multi-Processor (SMP) servers and the RS/6000 Scalable
POWERparallel Systems (RS/6000 SP) in a no single point of failure server
configuration. The minimum configuration and sizing of each system CPU is
highly dependent on the user's application and data requirements.
Nevertheless, systems with 32 MB of main storage and 1 GB of disk storage
would be practical, minimum configurations.
Almost any model of the RISC System/6000 POWERserver family can be
included in an HACMP environment and new models continue to be added to
the list. The following table gives you an overview of the currently supported
© Copyright IBM Corp. 1999
7
Download from Www.Somanuals.com. All Manuals Search And Download.
RISC System/6000 models as nodes in an HACMP 4.1 for AIX, HACMP 4.2
for AIX, or HACMP 4.3 for AIX cluster.
Table 3. Hardware Requirements for the Different HACMP Versions
HACMP Version
4.1
4.2
4.3
4.2/ES
4.3/ES
1
7009 Mod. CXX
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
no
no
no
no
no
no
no
no
no
no
yes
yes
1
7011 Mod. 2XX
yes
1
7012 Mod. 3XX and GXX
7013 Mod. 5XX and JXX
7015 Mod. 9XX and RXX
7017 Mod. S7X
yes
1
yes
1
yes
1
yes
1
7024 Mod. EXX
yes
1
7025 Mod. FXX
yes
1
7026 Mod. HXX
yes
1
7043 Mod. 43P, 260
9076 RS/6000 SP
yes
1
yes
1
AIX 4.3.2 required
For a detailed description of system models supported by HACMP/6000 and
HACMP/ES, you should refer to the current Announcement Letters for
HACMP/6000 and HACMP/ES.
HACMP/ES 4.3 further enhances cluster design flexibility even further by
including support for the RISC System/6000 family of machines and the
Compact Server C20. Since the introduction of HACMP 4.1 for AIX, you have
been able to mix uniprocessor and multiprocessor machines in a single
cluster. Even a mixture of “normal” RS/6000 machines and RS/6000 SP
nodes is possible.
2.1.2 Cluster Node Considerations
It is important to understand that selecting the system components for a
cluster requires careful consideration of factors and information that may not
be considered in the selection of equipment for a single-system environment.
In this section, we will offer some guidelines to assist you in choosing and
sizing appropriate machine models to build your clusters.
8
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Much of the decision centers around the following areas:
• Processor capacity
• Application requirements
• Anticipated growth requirements
• I/O slot requirements
These paradigms are certainly not new ones, and are also important
considerations when choosing a processor for a single-system environment.
However, when designing a cluster, you must carefully consider the
requirements of the cluster as a total entity. This includes understanding
system capacity requirements of other nodes in the cluster beyond the
requirements of each system's prescribed normal load. You must consider
the required performance of the solution during and after failover, when a
surviving node has to add the workload of a failed node to its own workload.
For example, in a two node cluster, where applications running on both nodes
are critical to the business, each of the two nodes functions as a backup for
the other, in a mutual takeover configuration. If a node is required to provide
failover support for all the applications on the other node, then its capacity
calculation needs to take that into account. Essentially, the choice of a model
depends on the requirements of highly available applications, not only in
terms of CPU cycles, but also of memory and possibly disk space.
Approximately 50 MB of disk storage is required for full installation of the
HACMP software.
A major consideration in the selection of models will be the number of I/O
expansion slots they provide. The model selected must have enough slots to
house the components required to remove single points of failure (SPOFs)
and provide the desired level of availability. A single point of failure is defined
as any single component in a cluster whose failure would cause a service to
become unavailable to end users. The more single points of failure you can
eliminate, the higher your level of availability will be. Typically, you need to
consider the number of slots required to support network adapters and disk
I/O adapters. Your slot configuration must provide for at least two network
adapters to provide adapter redundancy for one service network. If your
system needs to be able to take over an IP address for more than one other
system in the cluster at a time, you will want to configure more standby
network adapters. A node can have up to seven standby adapters for each
network it connects to. Again, if that is your requirement, you will need to
select models as nodes where the number of slots will accomodate the
requirement.
Cluster Planning
9
Download from Www.Somanuals.com. All Manuals Search And Download.
Your slot configuration must also allow for the disk I/O adapters you need to
support the cluster’s shared disk (volume group) configuration. If you intend
to use disk mirroring for shared volume groups, which is strongly
recommended, then you will need to use slots for additional disk I/O
adapters, providing I/O adapter redundancy across separate buses.
The following table tells you the number of additional adapters you can put
into the different RS/6000 models. Ethernet environments can sometimes
make use of the integrated ethernet port provided by some models. No such
feature is available for token-ring, FDDI or ATM; you must use an I/O slot to
provide token-ring adapter redundancy.
Table 4. Number of Adapter Slots in Each Model
RS/6000 Model
7006
Number of Slots
4 x MCA
Integrated Ethernet Port
yes
no
7009 C10, C20
7012 Mod. 3XX and GXX
7013 Mod. 5XX
7013 Mod. JXX
4x PCI
4 x MCA
yes
no
7 x MCA
6 x MCA, 14 x MCA no
with expansion unit J01
7015 Mod. R10, R20, R21
7015 Mod. R30, R40, R50
7017 Mod. S7X
8 x MCA
16 x MCA
52 x PCI
no
no
no
7024 EXX
5 x PCI, 1 x PCI/ISA 2 x no
ISA
7025 F50
6 x PCI, 2 x ISA/PCI
6 x PCI, 2 x ISA/PCI
3 x PCI, 2 x ISA/PCI
4 x MCA
yes
7026 Mod. H50
7043Mod.
yes
yes
yes
no
9076 thin node
9076 wide node
9076 high node
9076 thin node (silver)
9076 wide node (silver)
7 x MCA
15 x MCA
no
1
1
2 x PCI
yes
10 x PCI
yes
1
The switch adapter is onboard and does not need an extra slot.
10
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
2.2 Cluster Networks
HACMP differentiates between two major types of networks: TCP/IP networks
and non-TCP/IP networks. HACMP utilizes both of them for exchanging
heartbeats. HACMP uses these heartbeats to diagnose failures in the cluster.
Non-TCP/IP networks are used to distinguish an actual hardware failure from
the failure of the TCP/IP software. If there were only TCP/IP networks being
used, and the TCP/IP software failed, causing heartbeats to stop, HACMP
could falsely diagnose a node failure when the node was really still
functioning. Since a non-TCP/IP network would continue working in this
event, the correct diagnosis could be made by HACMP. In general, all
networks are also used for verification, synchronization, communication and
triggering events between nodes. Of course, TCP/IP networks are used for
communication with client machines as well.
At the time of publication, the HACMP/ES Version 4.3 product does not use
non-TCP/IP networks for node-to-node communications in triggering,
synchronizing, and executing event reactions. This can be an issue if you are
configuring a cluster with only one TCP/IP network. This limitation of
HACMP/ES is planned to be removed in a future release. You would be
advised to check on the status of this issue if you are planning a new
installation, and to plan your cluster networks accordingly.
2.2.1 TCP/IP Networks
The following sections describe supported TCP/IP network types and network
considerations.
2.2.1.1 Supported TCP/IP Network Types
Basically every adapter that is capable of running the TCP/IP protocol is a
supported HACMP network type. There are some special considerations for
certain types of adapters however. The following gives a brief overview on the
supported adapters and their special considerations.
Below is a list of TCP/IP network types as you will find them at the
configuration time of an adapter for HACMP. You will find the non-TCP/IP
• Generic IP
• ATM
• Ethernet
• FCS
Cluster Planning 11
Download from Www.Somanuals.com. All Manuals Search And Download.
• FDDI
• SP Switch
• SLIP
• SOCC
• Token-Ring
As an independent, layered component of AIX, the HACMP for AIX software
works with most TCP/IP-based networks. HACMP for AIX has been tested
with standard Ethernet interfaces (en*) but not with IEEE 802.3 Ethernet
interfaces (et*), where * reflects the interface number. HACMP for AIX also
has been tested with Token-Ring and Fiber Distributed Data Interchange
(FDDI) networks, with IBM Serial Optical Channel Converter (SOCC), Serial
Line Internet Protocol (SLIP), and Asynchronous Transfer Mode (ATM)
point-to-point connections.
Note
ATM and SP Switch networks are special cases of point-to-point, private
networks that can connect clients
The HACMP for AIX software supports a maximum of 32 networks per cluster
and 24 TCP/IP network adapters on each node. These numbers provide a
great deal of flexibility in designing a network configuration. The network
design affects the degree of system availability in that the more
communication paths that connect clustered nodes and clients, the greater
the degree of network availability.
2.2.1.2 Special Network Considerations
Each type of interface has different characteristics concerning speed, MAC
addresses, ARP, and so on. In case there is a limitation you will have to work
around, you need to be aware of the characteristics of the adapters you plan
to use. In the next paragraphs, we summarize some of the considerations
that are known.
Hardware Address Swapping is one issue. If you enable HACMP to put one
address on another adapter, it would need something like a boot and a
service address for IPAT, but on the hardware layer. So, in addition to the
manufacturers burnt-in address, there has to be an alternate address
configured.
The speed of the network can be another issue. Your application may have
special network throughput requirements that must be taken into account.
12
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Network types also differentiate themselves in the maximum distance they
allow between adapters, and in the maximum number of adapters allowed on
a physical network.
• Ethernet supports 10 and 100 Mbps currently, and supports hardware
address swapping. Alternate hardware addresses should be in the form
xxxxxxxxxxyy, where xxxxxxxxxx is replaced with the first five pairs of digits
of the original burned-in MAC address and yy can be chosen freely. There
is a limit of 29 adapters on one physical network, unless a network
repeater is used.
• Token-Ring supports 4 or 16 Mbps, but 4 Mbps is very rarely used now. It
also supports hardware address swapping, but here the convention is to
use 42 as the first two characters of the alternate address, since this
indicates that it is a locally set address.
• FDDI is a 100 Mbps optical LAN interface, that supports hardware address
takeover as well. For FDDI adapters you should leave the last six digits of
the burned-in address as they are, and use a 4, 5, 6, or 7 as the first digit
of the rest. FDDI can connect as many as 500 stations with a maximum
link-to-link distance of two kilometers and a total LAN circumference of
100 kilometers.
• ATM is a point-to-point connection network. It currently supports the OC3
and the OC12 standard, which is 155 Mbps or 625 Mbps. You cannot use
hardware address swapping with ATM. ATM doesn’t support broadcasts,
so it must be configured as a private network to HACMP. However, if you
are using LAN Emulation on an existing ATM network, you can use the
emulated ethernet or Token-Ring interfaces just as if they were real ones,
except that you cannot use hardware address swapping.
• FCS is a fiber channel network, currently available as two adapters for
either MCA or PCI technology. The Fibre Channel Adapter /1063-MCA,
runs up to 1063 Mb/second, and the Gigabit Fibre Channel Adapter for
PCI Bus (#6227), announced on October 5th 1998, will run with 100 MBps.
Both of them support TCP/IP, but not hardware address swapping.
•SLIP runs at up to 38400 bps. Since it is a point-to-point connection and
very slow, it is rarely used as an HACMP network. An HACMP cluster is
much more likely to use the serial port as a non-TCP/IP connection. See
below for details.
• SOCC is a fast optical connection, again point-to-point. This is an optical
line with a serial protocol running on it. However, the SOCC Adapter
(Feature 2860) has been withdrawn from marketing for some years now.
Some models, like 7013 5xx, offer SOCC as an option onboard, but these
are rarely used today.
Cluster Planning 13
Download from Www.Somanuals.com. All Manuals Search And Download.
• SP Switch is a high-speed packet switching network, running on the
RS/6000 SP system only. It runs bidirectionally up to 80 MBps, which adds
up to 160 MBps of capacity per adapter. This is node-to-node
communication and can be done in parallel between every pair of nodes
inside an SP. The SP Switch network has to be defined as a private
Network, and ARP must be enabled. This network is restricted to one
adapter per node, thus, it has to be considered as a Single Point Of
Failure. Therefore, it is strongly recommended to use AIX Error
Notification to propagate a switch failure into a node failure when
appropriate. As there is only one adapter per node, HACMP uses the
ifconfig alias addresses for IPAT on the switch; so, a standby address is
not necessary and, therefore, not used on the switch network. Hardware
address swapping also is not supported on the SP Switch.
For IP Address Takeover (IPAT), in general, there are two adapters per
cluster node and network recommended in order to eliminate single points of
failure. The only exception to this rule is the SP Switch because of hardware
limitations.
2.2.2 Non-TCPIP Networks
Non-TCP/IP networks in HACMP are used as an independent path for
exchanging messages or heartbeats between cluster nodes. In case of an IP
subsystem failure, HACMP can still differentiate between a network failure
and a node failure when an independent path is available and functional.
Below is a short description of the three currently available non-TCP/IP
network types and their characteristics. Even though HACMP works without
one, it is strongly recommended that you use at least one non-TCP/IP
connection between the cluster nodes.
2.2.2.1 Supported Non-TCP/IP Network Types
Currently HACMP supports the following types of networks for non-TCP/IP
heartbeat exchange between cluster nodes:
• Serial (RS232)
• Target-mode SCSI
• Target-mode SSA
All of them must be configured as Network Type: serial in the HACMP
definitions.
14
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
2.2.2.2 Special Considerations
As for TCP/IP networks, there are a number of restrictions on non-TCP/IP
networks. These are explained for the three different types in more detail
below.
Serial (RS232)
A serial (RS232) network needs at least one available serial port per cluster
node. In case of a cluster consisting of more than two nodes, a ring of nodes
is established through serial connections, which requires two serial ports per
node. Table 3 shows a list of possible cluster nodes and the number of native
serial ports for each:
Table 5. Number of Available Serial Ports in Each Model.
RS/6000 Model
7006
Number of Serial Ports Available
1
1
1
7009 C10, C20
1
7012 Mod. 3XX and GXX
7013 Mod. 5XX
2
2
3
3
3
7013 Mod. JXX
7015 Mod. R10, R20, R21
7015 Mod. R30, R40, R50
7013,7015,7017 Mod. S7X
7024 EXX
2
0
1
1
7025 F50
2
3
2
7026 Mod. H50
7043Mod.
3
9076 thin node
2
3
9076 wide node
9076 high node
2
3
3
3
9076 thin node (silver)
9076 wide node (silver)
2
3
2
1
serial port can be multiplexed through a dual-port cable, thus offering two
ports
Cluster Planning 15
Download from Www.Somanuals.com. All Manuals Search And Download.
2
3
a PCI Multiport Async Card is required in an S7X model, no native ports
only one serial port available for customer use, i.e. HACMP
In case the number of native serial ports doesn’t match your HACMP cluster
configuration needs, you can extend it by adding an eight-port asynchronous
adapter, thus reducing the number of available MCA slots, or the
corresponding PCI Multiport Async Card for PCI Machines, like the S7X
model.
Target-mode SCSI
Another possibility for a non-TCP/IP network is a target mode SCSI
connection. Whenever you make use of a shared SCSI device, you can also
use the SCSI bus for exchanging heartbeats.
Target Mode SCSI is only supported with SCSI-2 Differential or SCSI-2
Differential Fast/Wide devices. SCSI/SE or SCSI-2/SE are not supported for
HACMP serial networks.
The recommendation is to not use more than 4 target mode SCSI networks in
a cluster.
Target-mode SSA
If you are using shared SSA devices, target mode SSA is the third possibility
for a serial network within HACMP. In order to use target-mode SSA, you
must use the Enhanced RAID-5 Adapter (#6215 or #6219), since these are
the only current adapters that support the Multi-Initiator Feature. The
microcode level of the adapter must be 1801 or higher.
2.3 Cluster Disks
This section describes the various choices you have in selecting the type of
shared disks to use in your cluster.
2.3.1 SSA Disks
The following is a brief description of SSA and the basic rules to follow when
designing SSA networks. For a full description of SSA and its functionality,
please read Monitoring and Managing IBM SSA Disk Subsystems,
SG24-5251.
SSA is a high-performance, serial interconnect technology used to connect
disk devices and host adapters. SSA is an open standard, and SSA
specifications have been approved by the SSA Industry Association and also
as an ANSI standard through the ANSI X3T10.1 subcommittee.
16
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
SSA subsystems are built up from loops of adapters and disks. A simple
example is shown in Figure 1.
SSA Architecture
MBp
20 M
HOST
M
l
l
l
l
l
High performance 80 MB/s interface
Loop architecture with up to 127 nodes per loop
Up to 25 m (82 ft) between SSA devices with copper cables
Up to 2.4 km (1.5 mi) between SSA devices with optical extender
Spatial reuse (multiple simultaneous transmissions)
Figure 1. Basic SSA Configuration
Here, a single adapter controls one SSA loop of eight disks. Data can be
transferred around the loop, in either direction, at 20 MBps. Consequently,
the peak transfer rate of the adapter is 80 MBps. The adapter contains two
SSA nodes and can support two SSA loops. Each disk drive also contains a
single SSA node. A node can be either an initiator or a target. An initiator
issues commands, while a target responds with data and status information.
The SSA nodes in the adapter are therefore initiators, while the SSA nodes in
the disk drives are targets.
There are two types of SSA Disk Subsystems for RISC System/6000
available:
• 7131 SSA Multi-Storage Tower Model 405
Cluster Planning 17
Download from Www.Somanuals.com. All Manuals Search And Download.
• 7133 Serial Storage Architecture (SSA) Disk Subsystem Models 010, 500,
020, 600, D40 and T40.
The 7133 models 010 and 500 were the first SSA products announced in
1995 with the revolutionary new Serial Storage Architecture. Some IBM
customers still use the Models 010 and 500, but these have been replaced by
7133 Model 020, and 7133 Model 600 respectively. More recently, in
November 1998, the models D40 and T40 were announced.
All 7133 Models have redundant power and cooling, which is hot-swappable.
The following tables give you more configuration information about the
different models:
Table 6. 7131-Model 405 SSA Multi-Storage Tower Specifications
Item
Specification
80 MB
Transfer rate SSA
interface
Configuration
2 to 5 disk drives (2.2 GB, 4.5 GB or 9.1 GB) per subsystem
Configuration range
4.4 to 11 GB (with 2.2 GB disk drives)
9.0 to 22.5 GB (With 4.5 GB disk drives)
18.2 to 45.5 GB (With 9.1 GB disk drives)
Supported RAID
levels
5
Supported adapters
Hot-swap disks
6214, 6216, 6217, 6218
Yes
Table 7. 7133 Models 010, 020, 500, 600, D40, T40 Specifications
Item
Specification
80 MB/s
Transfer rate SSA
interface
Configuration
4 to 16 disks
- 1.1 GB, 2.2 GB, 4.5 GB, for Models 10, 20, 500, and 600
- 9.1 GB for Models 20, 600, D40 and T40
- With 1.1 GB disk drives you must have 8 to 16 disks)
Configuration range
8.8 to 17.6 GB (with 1.1 GB disks)
8.8 to 35.2 GB (with 2.2 GB disks)
18 to 72 GB (with 4.5 GB disks)
36.4 to 145.6 GB (with 9.1 GB disks)
72.8 to 291.2 GB (with 18.2 GB disks)
18
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Item
Specification
Supported RAID level
Supported adapters
Hot-swappable disk
5
all
Yes (and hot-swappable, redundant power and cooling)
2.3.1.1 Disk Capacities
characteristics.
Table 8. SSA Disks
Name
Capacities (GB)
Buffer size (KB)
Maximum Transfer rate
(MBps)
Starfire 1100
Starfire 2200
Starfire 4320
Scorpion 4500
Scorpion 9100
Sailfin 9100
Thresher 9100
Ultrastar
1.1
0
20
2.2
0
20
4.5
512
512
512
1024
1024
4096
20
4.5
80
9.1
160
160
160
160
9.1
9.1
9.1, 18.2
2.3.1.2 Supported and Non-Supported Adapters
characteristics.
Table 9. SSA Adapters
Feature Code Adapter
Label
Bus
Adapter Description
Number
Adapters per Raid Types
Loop
of Hardware
6214
6215
6216
6217
6218
4-D
4-N
4-G
4-I
MCA
PCI
Classic
2
8
8
1
1
n/a
5
1
Enhanced RAID-5
Enhanced
RAID-5
MCA
MCA
PCI
n/a
5
4-J
RAID-5
5
Cluster Planning 19
Download from Www.Somanuals.com. All Manuals Search And Download.
Feature Code Adapter
Label
Bus
Adapter Description
Enhanced RAID-5
Number
Adapters per Raid Types
Loop
of Hardware
1
6219
4-M
MCA
8
5
1See 2.3.1.3, “Rules for SSA Loops” on page 20 for more information.
The following rules apply to SSA Adapters:
• You cannot have more than four adapters in a single system.
• The MCA SSA 4-Port RAID Adapter (FC 6217) and PCI SSA 4-Port RAID
Adapter (FC 6218) are not useful for HACMP, because only one can be in
a loop.
• Only the PCI Multi Initiator/RAID Adapter (FC 6215) and the MCA Multi
Initiator/RAID EL Adapter (FC 6219) support target mode SSA (for more
information about target mode SSA see 3.2.2, “Non TCP/IP Networks” on
page 63).
2.3.1.3 Rules for SSA Loops
The following rules must be followed when configuring and connecting SSA
loops:
• Each SSA loop must be connected to a valid pair of connectors on the
SSA adapter (that is, either Connectors A1 and A2, or Connectors B1 and
B2).
• Only one of the two pairs of connectors on an adapter card can be
connected in a single SSA loop.
• A maximum of 48 devices can be connected in a single SSA loop.
• A maximum of two adapters can be connected in a particular loop if one
adapter is an SSA 4-Port adapter, Feature 6214.
• A maximum of eight adapters can be connected in a particular loop if all
the adapters are Enhanced SSA 4-Port Adapters, Feature 6216.
• A maximum of two SSA adapters, both connected in the same SSA loop,
can be installed in the same system.
For SSA loops that include an SSA Four-Port RAID adapter (Feature 6217) or
a PCI SSA Four-Port RAID adapter (Feature 6218), the following rules apply:
• Each SSA loop must be connected to a valid pair of connectors on the
SSA adapter (that is, either Connectors A1 and A2, or Connectors B1 and
B2).
20
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• A maximum of 48 devices can be connected in a particular SSA loop.
• Only one pair of adapter connectors can be connected in a particular SSA
loop.
• Member disk drives of an array can be on either SSA loop.
For SSA loops that include a Micro Channel Enhanced SSA
Multi-initiator/RAID EL adapter, Feature 6215 or a PCI SSA
Multi-initiator/RAID EL adapter, Feature 6219, the following rules apply:
• Each SSA loop must be connected to a valid pair of connectors on the
SSA adapter (that is, either Connectors A1 and A2, or Connectors B1 and
B2).
• A maximum of eight adapters can be connected in a particular loop if none
of the disk drives in the loops are array disk drives and none of them is
configured for fast-write operations. The adapters can be up to eight
Micro Channel Enhanced SSA Multi-initiator/RAID EL Adapters, up to
eight PCI Multi-initiator/RAID EL Adapters, or a mixture of the two types.
• A maximum of two adapters can be connected in a particular loop if one or
more of the disk drives in the loop are array disk drives that are not
configured for fast-write operations. The adapters can be two Micro
Channel Enhanced SSA Multi-initiator/RAID EL Adapters, two PCI
Multi-initiator/RAID EL Adapters, or one adapter of each type.
• Only one Micro Channel Enhanced SSA Multi-initiator/RAID EL Adapter or
PCI SSA Multi-initiator/RAID EL Adapter can be connected in a particular
loop if any disk drives in the loops are members of a RAID-5 array, and are
configured for fast-write operations.
• All member disk drives of an array must be on the same SSA loop.
• A maximum of 48 devices can be connected in a particular SSA loop.
• Only one pair of adapter connectors can be connected in a particular loop.
• When an SSA adapter is connected to two SSA loops, and each loop is
connected to a second adapter, both adapters must be connected to both
loops.
For the IBM 7190-100 SCSI to SSA converter, the following rules apply:
• There can be up to 48 disk drives per loop.
• There can be up to four IBM 7190-100 attached to any one SSA loop.
Cluster Planning 21
Download from Www.Somanuals.com. All Manuals Search And Download.
2.3.1.4 RAID vs. Non-RAID
RAID Technology
RAID is an acronym for Redundant Array of Independent Disks. Disk arrays
are groups of disk drives that work together to achieve higher data-transfer
and I/O rates than those provided by single large drives.
Arrays can also provide data redundancy so that no data is lost if a single
drive (physical disk) in the array should fail. Depending on the RAID level,
data is either mirrored or striped. The following gives you more information
about the different RAID levels.
RAID Level 0
RAID 0 is also known as data striping. Conventionally, a file is written out
sequentially to a single disk. With striping, the information is split into chunks
(fixed amounts of data usually called blocks) and the chunks are written to (or
read from) a series of disks in parallel. There are two main performance
advantages to this:
1. Data transfer rates are higher for sequential operations due to the
overlapping of multiple I/O streams.
2. Random access throughput is higher because access pattern skew is
eliminated due to the distribution of the data. This means that with data
distributed evenly across a number of disks, random accesses will most
likely find the required information spread across multiple disks and thus
benefit from the increased throughput of more than one drive.
RAID 0 is only designed to increase performance. There is no redundancy; so
any disk failures will require reloading from backups.
RAID Level 1
RAID 1 is also known as disk mirroring. In this implementation, identical
copies of each chunk of data are kept on separate disks, or more commonly,
each disk has a twin that contains an exact replica (or mirror image) of the
information. If any disk in the array fails, then the mirrored twin can take over.
Read performance can be enhanced because the disk with its actuator
closest to the required data is always used, thereby minimizing seek times.
The response time for writes can be somewhat slower than for a single disk,
depending on the write policy; the writes can either be executed in parallel for
speed or serially for safety.
RAID Level 1 has data redundancy, but data should be regularly backed up
on the array. This is the only way to recover data in the event that a file or
directory is accidentally deleted.
22
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
RAID Levels 2 and 3
RAID 2 and RAID 3 are parallel process array mechanisms, where all drives
in the array operate in unison. Similar to data striping, information to be
written to disk is split into chunks (a fixed amount of data), and each chunk is
written out to the same physical position on separate disks (in parallel). When
a read occurs, simultaneous requests for the data can be sent to each disk.
This architecture requires parity information to be written for each stripe of
data; the difference between RAID 2 and RAID 3 is that RAID 2 can utilize
multiple disk drives for parity, while RAID 3 can use only one. If a drive should
fail, the system can reconstruct the missing data from the parity and
remaining drives.
Performance is very good for large amounts of data but poor for small
requests since every drive is always involved, and there can be no
overlapped or independent operation.
RAID Level 4
RAID 4 addresses some of the disadvantages of RAID 3 by using larger
chunks of data and striping the data across all of the drives except the one
reserved for parity. Using disk striping means that I/O requests need only
reference the drive that the required data is actually on. This means that
simultaneous, as well as independent reads, are possible. Write requests,
however, require a read/modify/update cycle that creates a bottleneck at the
single parity drive. Each stripe must be read, the new data inserted and the
new parity then calculated before writing the stripe back to the disk. The
parity disk is then updated with the new parity, but cannot be used for other
writes until this has completed. This bottleneck means that RAID 4 is not
used as often as RAID 5, which implements the same process but without the
bottleneck. RAID 5 is discussed in the next section.
RAID Level 5
RAID 5, as has been mentioned, is very similar to RAID 4. The difference is
that the parity information is distributed across the same disks used for the
data, thereby eliminating the bottleneck. Parity data is never stored on the
same drive as the chunks that it protects. This means that concurrent read
and write operations can now be performed, and there are performance
increases due to the availability of an extra disk (the disk previously used for
parity). There are other enhancements possible to further increase data
transfer rates, such as caching simultaneous reads from the disks and
transferring that information while reading the next blocks. This can generate
data transfer rates that approach the adapter speed.
Cluster Planning 23
Download from Www.Somanuals.com. All Manuals Search And Download.
As with RAID 3, in the event of disk failure, the information can be rebuilt from
the remaining drives. RAID level 5 array also uses parity information, though
it is still important to make regular backups of the data in the array. RAID level
5 stripes data across all of the drives in the array, one segment at a time (a
segment can contain multiple blocks). In an array with n drives, a stripe
consists of data segments written to n-1 of the drives and a parity segment
written to the nth drive. This mechanism also means that not all of the disk
space is available for data. For example, in an array with five 2 GB disks,
although the total storage is 10 GB, only 8 GB are available for data.
The advantages and disadvantages of the various RAID levels are
summarized in the following table:
Table 10. The Advantages and Disadvantages of the Different RAID Levels
RAID Level
Availability
Mechanism
Capacity
Performance
Cost
0
1
3
5
none
100%
50%
80%
80%
high
medium
high
mirroring
parity
medium/high
medium
medium
medium
medium
parity
RAID on the 7133 Disk Subsystem
The only RAID level supported by the 7133 SSA disk subsystem is RAID 5.
RAID 0 and RAID 1 can be achieved with the striping and mirroring facility of
the Logical Volume Manager (LVM).
RAID 0 does not provide data redundancy, so it is not recommended for use
with HACMP, because the shared disks would be a single point of failure. The
possible configurations to use with the 7133 SSA disk subsystem are RAID 1
(mirroring) or RAID 5. Consider the following points before you make your
decision:
• Mirroring is more expensive than RAID, but it provides higher data
redundancy. Even if more than one disk fails, you may still have access to
all of your data. In a RAID, more than one broken disk means that the data
are lost.
• The SSA loop can include a maximum of two SSA adapters if you use
RAID. So, if you want to connect more than two nodes into the loop,
mirroring is the way to go.
• A RAID array can consist of three to 16 disks.
24
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Array member drives and spares must be on same loop (cannot span A
and B loops) on the adapter.
• You cannot boot (ipl) from a RAID.
2.3.1.5 Advantages
Because SSA allows SCSI-2 mapping, all functions associated with initiators,
targets, and logical units are translatable. Therefore, SSA can use the same
command descriptor blocks, status codes, command queuing, and all other
aspects of current SCSI systems. The effect of this is to make the type of disk
subsystem transparent to the application. No porting of applications is
required to move from traditional SCSI I/O subsystems to high-performance
SSA. SSA and SCSI I/O systems can coexist on the same host running the
same applications.
The advantages of SSA are summarized as follows:
• Dual paths to devices.
• Simplified cabling - cheaper, smaller cables and connectors, no separate
terminators.
• Faster interconnect technology.
• Not an arbitrated system.
• Full duplex, frame multiplexed serial links.
• 40 MBps total per port, resulting in 80 MBps total per node, and 160 MBps
total per adapter.
• Concurrent access to disks.
• Hot-pluggable cables and disks.
• Very high capacity per adapter - up to 127 devices per loop, although most
adapter implementations limit this. For example, current IBM SSA
adapters provide 96 disks per Micro Channel or PCI slot.
• Distance between devices of up to 25 meters with copper cables, 10km
with optical links.
• Auto-configuring - no manual address allocation.
• SSA is an open standard.
• SSA switches can be introduced to produce even greater fan-out and
more complex topologies.
Cluster Planning 25
Download from Www.Somanuals.com. All Manuals Search And Download.
2.3.2 SCSI Disks
After the announcement of the 7133 SSA Disk Subsystems, the SCSI Disk
subsystems became less common in HACMP clusters. However, the 7135
RAIDiant Array (Model 110 and 210) and other SCSI Subsystems are still in
use at many customer sites. We will not describe other SCSI Subsystems
such as 9334 External SCSI Disk Storage. See the appropriate
documentation if you need information about these SCSI Subsystems.
The 7135 RAIDiant Array is offered with a range of features, with a maximum
capacity of 135 GB (RAID 0) or 108 GB (RAID-5) in a single unit, and uses
the 4.5 GB disk drive modules. The array enclosure can be integrated into a
RISC System/6000 system rack, or into a deskside mini-rack. It can attach to
multiple systems through a SCSI-2 Differential 8-bit or 16-bit bus.
2.3.2.1 Capacities
Disks
There are four disk sizes available for the 7135 RAIDiant Array Models 110
and 210:
• 1.3 GB
• 2.0 GB
• 2.2 GB (only supported by Dual Active Software)
• 4.5 GB (only supported by Dual Active Software)
Subsystems
The 7135-110/210 can contain 15 Disks (max. 67.5 GB) in the base
configuration and 30 Disks (max. 135 GB) in an extended configuration.You
can for example only use the full 135 GB storage space for data if you
configure the 7135 with RAID level 0. When using RAID level 5, only 108 GB
of the 135 GB are available for data storage.
2.3.2.2 How Many in a String?
HACMP supports a maximum of two 7135s on a shared SCSI bus. This is
because of cable length restrictions.
2.3.2.3 Supported SCSI Adapters
The SCSI adapters that can be used to connect RAID subsystems on a
shared SCSI bus in an HACMP cluster are:
• SCSI-2 Differential Controller (MCA, FC: 2420, Adapter Label: 4-2)
• SCSI-2 Differential Fast/Wide Adapter/A (MCA, FC: 2416, Adapter Label:
4-6)
26
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Enhanced SCSI-2 Differential Fast/Wide Adapter/A (MCA, FC: 2412,
Adapter Label: 4-C); not usable with 7135-110
• SCSI-2 Fast/Wide Differential Adapter (PCI, FC: 6209, Adapter Label:
4-B)
• DE Ultra SCSI Adapter (PCI, FC: 6207, Adapter Label: 4-L); not usable
with 7135-110
2.3.2.4 Advantages - Disadvantages
The 7135 RAIDiant Array incorporates the following high availability features:
• Support for RAID-1, RAID-3 (Model 110 only) and RAID-5
You can run any combination of RAID levels in a single 7135 subsystem.
Each LUN can run its own RAID level.
• Multiple Logical Unit (LUN) support
The RAID controller takes up only one SCSI ID on the external bus. The
internal disks are grouped into logical units (LUNs). The array will support
up to six LUNs, each of which appears to AIX as a single hdisk device.
Since each of these LUNs can be configured into separate volume groups,
different parts of the subsystem can be logically attached to different
systems at any one time.
• Redundant Power Supply
Redundant power supplies provide alternative sources of power. If one
supply fails, power is automatically supplied by the other.
• Redundant Cooling
Extra cooling fans are built into the RAIDiant Array to safeguard against
fan failure.
• Concurrent Maintenance
Power supplies, cooling fans, and failed disk drives can be replaced
without the need to take the array offline or to power it down.
• Optional Second Array Controller
This allows the array subsystem to be configured with no single point of
failure. Under the control of the system software, the machine can be
configured in Dual Active mode, so that each controller controls the
operation of specific sets of drives. In the event of failure of either
controller, all I/O activity is switched to the remaining active controller.
In the last few years, the 7133 SSA Subsystems have become more popular
than 7135 RAIDiant Systems due to better technology. IBM decided to
Cluster Planning 27
Download from Www.Somanuals.com. All Manuals Search And Download.
withdraw the 7135 RAIDiant Systems from marketing because it is equally
possible to configure RAID on the SSA Subsystems.
2.4 Resource Planning
HACMP provides a highly available environment by identifying a set of
cluster-wide resources essential to uninterrupted processing, and then
defining relationships among nodes that ensure these resources are
available to client processes.
When a cluster node fails or detaches from the cluster for a scheduled
outage, the Cluster Manager redistributes its resources among any number of
the surviving nodes.
HACMP considers the following as resource types:
• Volume Groups
• Disks
• File Systems
• File Systems to be NFS mounted
• File Systems to be NFS exported
• Service IP addresses
• Applications
The following paragraphs will tell you what to consider when configuring
resources to accomplish the following:
• IP Address Takeover
• Shared LVM Components
• NFS Exports
and the options you have when combining these resources to a resource
group.
2.4.1 Resource Group Options
Each resource in a cluster is defined as part of a resource group. This allows
you to combine related resources that need to be together to provide a
particular service. A resource group also includes the list of nodes that can
acquire those resources and serve them to clients.
A resource group is defined as one of three types:
28
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Cascading
• Rotating
• Concurrent
Each of these types describes a different set of relationships between nodes
in the cluster, and a different set of behaviors upon nodes entering and
leaving the cluster.
Cascading Resource Groups: All nodes in a cascading resource group are
assigned priorities for that resource group. These nodes are said to be part of
that group's resource chain. In a cascading resource group, the set of
resources cascades up or down to the highest priority node active in the
cluster. When a node that is serving the resources fails, the surviving node
with the highest priority takes over the resources.
A parameter called Inactive Takeover decides which node takes the
cascading resources when the nodes join the cluster for the first time. If this
parameter is set to true, the first node in a group's resource chain to join the
cluster acquires all the resources in the resource group. As successive nodes
join the resource group, the resources cascade up to any node with a higher
priority that joins the cluster. If this parameter is set to false, the first node in a
group's resource chain to join the cluster acquires all the resources in the
resource group only if it is the node with the highest priority for that group. If
the first node to join does not acquire the resource group, the second node in
the group's resource chain to join acquires the resource group, if it has a
higher priority than the node already active. As successive nodes join, the
resource group cascades to the active node with the highest priority for the
group. The default is false.
Member nodes of a cascading resource chain always release a resource
group to a reintegrating node with a higher priority.
Rotating Resource Groups: A rotating resource group is associated with a
group of nodes, rather than a particular node. A node can be in possession of
a maximum of one rotating resource group per network.
As participating nodes join the cluster for the first time, they acquire the first
available rotating resource group per network until all the groups are
acquired. The remaining nodes maintain a standby role.
When a node holding a rotating resource group leaves the cluster, either
because of a failure or gracefully while specifying the takeover option, the
node with the highest priority and available connectivity takes over. Upon
Cluster Planning 29
Download from Www.Somanuals.com. All Manuals Search And Download.
reintegration, a node remains as a standby and does not take back any of the
resources that it had initially served.
Concurrent Resource Groups: A concurrent resource group may be shared
simultaneously by multiple nodes. The resources that can be part of a
concurrent resource group are limited to volume groups with raw logical
volumes, raw disks, and application servers.
When a node fails, there is no takeover involved for concurrent resources.
Upon reintegration, a node again accesses the resources simultaneously with
the other nodes.
The Cluster Manager makes the following assumptions about the acquisition
of resource groups:
Cascading
The active node with the highest priority controls the
resource group.
Concurrent
Rotating
All active nodes have access to the resource group.
The node with the rotating resource group’s associated
service IP address controls the resource group.
2.4.2 Shared LVM Components
The first distinction that you need to make while designing a cluster is
whether you need a non-concurrent or a concurrent shared disk access
environment.
2.4.2.1 Non-Concurrent Disk Access Configurations
The possible non-concurrent disk access configurations are:
• Hot-Standby
• Rotating Standby
• Mutual Takeover
• Third-Party Takeover
Hot-Standby Configuration
Figure 2 illustrates a two node cluster in a hot-standby configuration.
30
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Figure 2. Hot-Standby Configuration
In this configuration, there is one cascading resource group consisting of the
four disks, hdisk1 to hdisk4, and their constituent volume groups and file
systems. Node 1 has a priority of 1 for this resource group while node 2 has a
priority of 2. During normal operations, node 1 provides all critical services to
end users. Node 2 may be idle or may be providing non-critical services, and
hence is referred to as a hot-standby node. When node 1 fails or has to leave
the cluster for a scheduled outage, node 2 acquires the resource group and
starts providing the critical services.
The advantage of this type of a configuration is that you can shift from a
single-system environment to an HACMP cluster at a low cost by adding a
less powerful processor. Of course, this assumes that you are willing to
accept a lower level of performance in a failover situation. This is a trade-off
that you will have to make between availability, performance, and cost.
Rotating Standby Configuration
This configuration is the same as the previous configuration except that the
resource groups used are rotating resource groups.
In the hot-standby configuration, when node 1 reintegrates into the cluster, it
takes back the resource group since it has the highest priority for it. This
implies a break in service to the end users during reintegration.
If the cluster is using rotating resource groups, reintegrating nodes do not
reacquire any of the resource groups. A failed node that recovers and rejoins
Cluster Planning 31
Download from Www.Somanuals.com. All Manuals Search And Download.
the cluster becomes a standby node. You must choose a rotating standby
configuration if you do not want a break in service during reintegration.
Since takeover nodes continue providing services until they have to leave the
cluster, you should configure your cluster with nodes of equal power. While
more expensive in terms of CPU hardware, a rotating standby configuration
gives you better availability and performance than a hot-standby
configuration.
Mutual Takeover Configuration
Figure 3 illustrates a two node cluster in a mutual takeover configuration.
Figure 3. Mutual Takeover Configuration
In this configuration, there are two cascading resource groups: A and B.
Resource group A consists of two disks, hdisk1 and hdisk3, and one volume
group, sharedvg. Resource group B consists of two disks, hdisk2 and hdisk4,
and one volume group, databasevg. Node 1 has priorities of 1 and 2 for
resource groups A and B respectively, while Node 2 has priorities of 1 and 2
for resource groups B and A respectively.
During normal operations, nodes 1 and 2 have control of resource groups A
and B respectively, and both provide critical services to end users. If either
node 1 or node 2 fails, or has to leave the cluster for a scheduled outage, the
surviving node acquires the failed node’s resource groups and continues to
provide the failed node’s critical services.
32
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
When a failed node reintegrates into the cluster, it takes back the resource
group for which it has the highest priority. Therefore, even in this
configuration, there is a break in service during reintegration. Of course, if
you look at it from the point of view of performance, this is the best thing to
do, since you have one node doing the work of two when any one of the
nodes is down.
Third-Party Takeover Configuration
Figure 4 illustrates a three node cluster in a third-party takeover
configuration.
Figure 4. Third-Party Takeover Configuration
This configuration can avoid the performance degradation that results from a
failover in the mutual takeover configuration.
Cluster Planning 33
Download from Www.Somanuals.com. All Manuals Search And Download.
Here the resource groups are the same as the ones in the mutual takeover
configuration. Also, similar to the previous configuration, nodes 1 and 2 each
have priorities of 1 for one of the resource groups, A or B. The only thing
different in this configuration is that there is a third node which has a priority
of 2 for both the resource groups.
During normal operations, node 3 is either idle or is providing non-critical
services. In the case of either node 1 or node 2 failing, node 3 takes over the
failed node’s resource groups and starts providing its services. When a failed
node rejoins the cluster, it reacquires the resource group for which it has the
highest priority.
So, in this configuration, you are protected against the failure of two nodes
and there is no performance degradation after the failure of one node.
2.4.2.2 Concurrent Disk Access Configurations
A concurrent disk access configuration usually has all its disk storage defined
as part of one concurrent resource group. The nodes associated with a
concurrent resource group have no priorities assigned to them.
If a 7135 RAIDiant Array Subsystem is used for storage, you can have a
maximum of four nodes concurrently accessing a set of storage resources. If
you are using the 7133 SSA Disk Subsystem, you can have up to eight nodes
concurrently accessing it.This is because of the physical characteristics of
SCSI versus SSA.
In the case of a node failure, a concurrent resource group is not explicitly
taken over by any other node, since it is already active on the other nodes.
However, in order to somewhat mask a node failure from the end users, you
should also have cascading resource groups, each containing the service IP
address for each node in the cluster. When a node fails, its service IP
address will be taken over by another node and users can continue to access
critical services at the same IP address that they were using before the node
failed.
2.4.3 IP Address Takeover
The goal of IP Address Takeover is to make the server’s service address
highly available and to give clients the possibility of always connecting to the
same IP address. In order to achieve this, you must do the following:
• Decide which types of networks and point-to-point connections to use in
types)
34
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Design the network topology
• Define a network mask for your site
• Define IP addresses (adapter identifiers) for each node’s service and
standby adapters.
• Define a boot address for each service adapter that can be taken over, if
you are using IP address takeover or rotating resources.
• Define an alternate hardware address for each service adapter that can
have its IP address taken over, if you are using hardware address
swapping.
2.4.3.1 Network Topology
The following sections cover topics of network topology.
Single Network
In a single-network setup, each node in the cluster is connected to only one
network and has only one service adapter available to clients. In this setup, a
service adapter on any of the nodes may fail, and a standby adapter will
acquire its IP address. The network itself, however, is a single point of failure.
The following figure shows a single-network configuration:
Client
Network
In the single-network setup, each node is connected to one network.
Each node has one service adapter and can have none, one, or more
standby adapters per public network.
Figure 5. Single-Network Setup
Cluster Planning 35
Download from Www.Somanuals.com. All Manuals Search And Download.
Dual Network
A dual-network setup has two separate networks for communication. Nodes
are connected to two networks, and each node has two service adapters
available to clients. If one network fails, the remaining network can still
function, connecting nodes and providing resource access to clients.
In some recovery situations, a node connected to two networks may route
network packets from one network to another. In normal cluster activity,
however, each network is separate—both logically and physically.
Keep in mind that a client, unless it is connected to more than one network, is
susceptible to network failure.
The following figure shows a dual-network setup:
Figure 6. Dual-Network Setup
Point-to-Point Connection
A point-to-point connection links two (neighboring) cluster nodes directly.
SOCC, SLIP, and ATM are point-to-point connection types. In HACMP
clusters of four or more nodes, however, use an SOCC line only as a private
network between neighboring nodes because it cannot guarantee cluster
communications with nodes other than its neighbors.
36
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The following diagram shows a cluster consisting of two nodes and a client. A
single public network connects the nodes and the client, and the nodes are
linked point-to-point by a private high-speed SOCC connection that provides
an alternate path for cluster and lock traffic should the public network fail.
Figure 7. A Point-to-Point Connection
2.4.3.2 Networks
Networks in an HACMP cluster are identified by name and attribute.
Network Name
The network name is a symbolic value that identifies a network in an HACMP
for AIX environment. Cluster processes use this information to determine
which adapters are connected to the same physical network. In most cases,
the network name is arbitrary, and it must be used consistently. If several
adapters share the same physical network, make sure that you use the same
network name when defining these adapters.
Network Attribute
A TCP/IP network’s attribute is either public or private.
Public
A public network connects from two to 32 nodes and allows clients to
monitor or access cluster nodes. Ethernet, Token-Ring, FDDI, and
Cluster Planning 37
Download from Www.Somanuals.com. All Manuals Search And Download.
SLIP are considered public networks. Note that a SLIP line, however,
does not provide client access.
Private
Serial
A private network provides communication between nodes only; it
typically does not allow client access. An SOCC line or an ATM
network are also private networks; however, an ATM network does
allow client connections and may contain standby adapters. If an
SP node is used as a client, the SP Switch network, although
private, can allow client access.
This network attribute is used for non TCP/IP networks (see 2.2.2,
2.4.3.3 Network Adapters
A network adapter (interface) connects a node to a network. A node typically
is configured with at least two network interfaces for each network to which it
connects: a service interface that handles cluster traffic, and one or more
standby interfaces. A service adapter must also have a boot address defined
for it if IP address takeover is enabled.
Adapters in an HACMP cluster have a label and a function (service, standby,
or boot). The maximum number of network interfaces per node is 24.
Adapter Label
A network adapter is identified by an adapter label. For TCP/IP networks, the
adapter label is the name in the /etc/hosts file associated with a specific IP
address. Thus, a single node can have several adapter labels and IP
addresses assigned to it. The adapter labels, however, should not be
confused with the “hostname”, of which there is only one per node.
Adapter Function
In the HACMP for AIX environment, each adapter has a specific function that
indicates the role it performs in the cluster. An adapter’s function can be
service, standby, or boot.
Service Adapter The service adapter is the primary connection between
the node and the network. A node has one service adapter
for each physical network to which it connects. The
service adapter is used for general TCP/IP traffic and is
the address the Cluster Information Program (Clinfo)
makes known to application programs that want to monitor
or use cluster services.
In configurations using rotating resources, the service
adapter on the standby node remains on its boot address
38
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
until it assumes the shared IP address. Consequently,
Clinfo makes known the boot address for this adapter.
In an HACMP for AIX environment on the RS/6000 SP, the
SP Ethernet adapters can be configured as service
adapters but should not be configured for IP address
takeover. For the SP switch network, service addresses
used for IP address takeover are ifconfig alias addresses
used on the css0 network.
Standby Adapter A standby adapter backs up a service adapter. If a service
adapter fails, the Cluster Manager swaps the standby
adapter’s address with the service adapter’s address.
Using a standby adapter eliminates a network adapter as
a single point of failure. A node can have no standby
adapter, or it can have from one to seven standby
adapters for each network to which it connects. Your
software configuration and hardware slot constraints
determine the actual number of standby adapters that a
node can support.
The standby adapter is configured on a different subnet
from any service adapters on the same system, and its
use should be reserved for HACMP only.
In an HACMP for AIX environment on the RS/6000 SP, for
an IP address takeover configuration using the SP switch,
standby adapters are not required.
Boot Adapter
IP address takeover is an AIX facility that allows one node
to acquire the network address of another node in the
cluster. To enable IP address takeover, a boot adapter
label (address) must be assigned to the service adapter
on each cluster node. Nodes use the boot label after a
system reboot and before the HACMP for AIX software is
started.
In an HACMP for AIX environment on the RS/6000 SP,
boot addresses used in the IP address for the switch
network takeover are ifconfig alias addresses used on
that css0 network.
When the HACMP for AIX software is started on a node,
the node’s service adapter is reconfigured to use the
Cluster Planning 39
Download from Www.Somanuals.com. All Manuals Search And Download.
service label (address) instead of the boot label. If the
node should fail, a takeover node acquires the failed
node’s service address on its standby adapter, thus
making the failure transparent to clients using that specific
service address.
During the reintegration of the failed node, which comes
up on its boot address, the takeover node will release the
service address it acquired from the failed node.
Afterwards, the reintegrating node will reconfigure its
adapter from the boot address to its reacquired service
address.
Consider the following scenario: Suppose that Node A
fails. Node B acquires Node A’s service address and
services client requests directed to that address. Later,
when Node A is restarted, it comes up on its boot address
and attempts to reintegrate into the cluster on its service
address by requesting that Node B release Node A’s
service address. When Node B releases the requested
address, Node A reclaims the address and reintegrates it
into the cluster. Reintegration, however, fails if Node A has
not been configured to boot using its boot address.
The boot address does not use a separate physical
adapter, but instead is a second name and IP address
associated with a service adapter. It must be on the same
subnetwork as the service adapter. All cluster nodes must
have this entry in the local /etc/hosts file and, if
applicable, in the nameserver configuration.
2.4.3.4 Defining Hardware Addresses
The hardware address swapping facility works in tandem with IP address
takeover. Hardware address swapping maintains the binding between an IP
address and a hardware address, which eliminates the need to flush the ARP
cache of clients after an IP address takeover. This facility, however, is
supported only for Ethernet, Token-Ring, and FDDI adapters. It does not work
with the SP Switch or ATM LAN emulation networks.
Note that hardware address swapping takes about 60 seconds on a
Token-Ring network, and up to 120 seconds on an FDDI network. These
periods are longer than the usual time it takes for the Cluster Manager to
detect a failure and take action.
40
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
If you do not use Hardware Address Takeover, the ARP cache of clients can
be updated by adding the clients’ IP addresses to the PING_CLIENT_LIST
variable in the /usr/sbin/cluster/etc/clinfo.rc file.
2.4.4 NFS Exports and NFS Mounts
There are two items concerning NFS when doing the configuration of a
Resource Group:
Filesystems to Export
File systems listed here will be NFS exported,
so they can be mounted by NFS client
systems or other nodes in the cluster.
Filesystems to NFS mount Filling in this field sets up what we call an NFS
cross mount. Any file system defined in this
field will be NFS mounted by all the
participating nodes, other than the node that is
currently holding the resource group. If the
node holding the resource group fails, the next
node to take over breaks its NFS mount for
this file system, and mounts the file system
itself as part of its takeover processing.
2.5 Application Planning
The central purpose for combining nodes in a cluster is to provide a highly
available environment for mission-critical applications. These applications
must remain available at all times in many organizations. For example, an
HACMP cluster could run a database server program that services client
applications. The clients send queries to the server program that responds to
their requests by accessing a database that is stored on a shared external
disk.
Planning for these applications requires that you be aware of their location
within the cluster, and that you provide a solution that enables them to be
handled correctly, in case a node should fail. In an HACMP for AIX cluster,
these critical applications can be a single point of failure. To ensure the
availability of these applications, the node configured to take over the
resources of the node leaving the cluster should also restart these
applications so that they remain available to client processes.
To put the application under HACMP control, you create an application server
cluster resource that associates a user-defined name with the names of
specially written scripts to start and stop the application. By defining an
application server, HACMP for AIX can start another instance of the
Cluster Planning 41
Download from Www.Somanuals.com. All Manuals Search And Download.
application on the takeover node when a fallover occurs. For more
information about creating application server resources, see the HACMP for
AIX, Version 4.3: Installation Guide, SC23-4278.
2.5.1 Performance Requirements
In order to plan your application’s needs, you must have a thorough
understanding of it. One part of that is to have The Application Planning
Worksheets, found in Appendix A of the HACMP for AIX Planning Guide,
SC23-4277, filled out.
Your applications have to be served correctly in an HACMP cluster
environment. Therefore, you need to know not only how they run on a single
uni- or multiprocessor machine, but also which resources are required by
them. How much disk space is required, what is the usual and critical load the
application puts on a server, and how users access the application are some
critical factors that will influence your decisions on how to plan the cluster.
Within an HACMP environment there are always a number of possible states
in which the cluster could be. Under normal conditions, the load is serviced
by a cluster node that was designed for this application’s needs. In case of a
failover, another node has to handle its own work plus the application it is
going to take over from a failing node. You can even plan one cluster node to
be the takeover node for multiple nodes; so, when any one of its primary
nodes fail, it has to take over its application and its load. Therefore, the
performance requirements of any cluster application have to be understood in
order to have the computing power available for mission-critical applications
in all possible cluster states.
2.5.2 Application Startup and Shutdown Routines
Highly available applications do not only have to come up at boot time, or
when someone is starting them up, but also when a critical resource fails and
has to be taken over by another cluster node. In this case, there have to be
robust scripts to both start up and shut down the application on the cluster
nodes. The startup script especially must be able to recover the application
from an abnormal termination, such as a power failure. You should verify that
it runs properly in a uniprocessor environment before including the HACMP
for AIX software.
42
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Note
Application start and stop scripts have to be available on the primary as
well as the takeover node. They are not transferred during synchronization;
so, the administrator of a cluster has to ensure that they are found in the
same path location, with the same permissions and in the same state, i.e.
changes have to be transferred manually.
2.5.3 Licensing Methods
Some vendors require a unique license for each processor that runs an
application, which means that you must license-protect the application by
incorporating processor-specific information into the application when it is
installed. As a result, it is possible that even though the HACMP for AIX
software processes a node failure correctly, it is unable to restart the
application on the failover node because of a restriction on the number of
licenses available within the cluster for that application. To avoid this
problem, make sure that you have a license for each system unit in the
cluster that may potentially run an application.
This can be done by “floating licenses”, where a license server is asked to
grant the permission to run an application on request, as well as “node-locked
licenses”, where each processor possibly running an application must have
the licensing files installed and configured.
2.5.4 Coexistence with other Applications
In case of a failover, a node might have to handle several applications
concurrently. This means the applications data or resources must not conflict
with each other. Again, the Application Worksheets can help in deciding
whether certain resources might conflict with others.
2.5.5 Critical/Non-Critical Prioritizations
Building a highly available environment for mission-critical applications also
forces the need to differentiate between the priorities of a number of
applications. Should a server node fail, it might be appropriate to shut down
another application, which is not as highly prioritized, in favor of the takeover
of the server node’s application. The applications running in a cluster have to
be clearly ordered and prioritized in order to decide what to do under these
circumstances.
Cluster Planning 43
Download from Www.Somanuals.com. All Manuals Search And Download.
2.6 Customization Planning
The Cluster Manager’s ability to recognize a specific series of events and
subevents permits a very flexible customization scheme. The HACMP for AIX
software provides an event customization facility that allows you to tailor
cluster event processing to your site.
2.6.1 Event Customization
As part of the planning process, you need to decide whether to customize
event processing. If the actions taken by the default scripts are sufficient for
your purposes, you do not need to do anything further to configure events
during the installation process.
If you decide to tailor event processing to your environment, it is strongly
recommended that you use the HACMP for AIX event customization facility
described in this chapter. If you tailor event processing, you must register
user-defined scripts with HACMP during the installation process. The HACMP
for AIX, Version 4.3: Installation Guide, SC23-4278 describes how to
configure event processing for a cluster.
You cannot define additional cluster events.
You can, however, define multiple pre- and post-events for each of the events
defined in the HACMPevent ODM class.
The event customization facility includes the following features:
• Event notification
• Pre- and post-event processing
• Event recovery and retry
2.6.1.1 Special Application Requirements
Some applications may have some special requirements that have to be
checked and ensured before or after a cluster event happens. In case of a
failover you can customize events through the definition of pre- and post-
events, to act according to your application’s needs. For example, an
application might want to reset a counter or unlock a user before it can be
started correctly on the failover node.
2.6.1.2 Event Notification
You can specify a notifycommand that sends mail to indicate that an event is
about to happen (or has just occurred), and that an event script succeeded or
failed. For example, a site may want to use a network_down notification
44
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
event to inform system administrators that traffic may have to be rerouted.
Afterwards, you can use a network_up notification event to inform system
administrators that traffic can again be serviced through the restored network.
2.6.1.3 Predictive Event Error Correction
You can specify a command that attempts to recover from an event script
failure. If the recovery command succeeds and the retry count for the event
script is greater than zero, the event script is rerun. You can also specify the
number of times to attempt to execute the recovery command.
For example, a recovery command can include the retry of unmounting a file
system after logging a user off and making sure no one was currently
accessing the file system.
If a condition that affects the processing of a given event on a cluster is
identified, such as a timing issue, you can insert a recovery command with a
retry count high enough to be sure to cover for the problem.
2.6.2 Error Notification
The AIX Error Notification facility detects errors that are logged to the AIX
error log, such as network and disk adapter failures, and triggers a predefined
response to the failure. It can even act on application failures, as long as they
are logged in the error log.
To implement error notification, you have to add an object to the Error
Notification object class in the ODM. This object clearly identifies what sort of
errors you are going to react to, and how.
By specifying the following in a file:
errnotify:
en_name = "Failuresample"
en_persistenceflg = 0
en_class = "H"
en_type = "PERM"
en_rclass = "disk"
en_method = "errpt -a -l $1 | mail -s ’Disk Error’ root"
and adding this to the errnotifyclass through the odmadd <filename>
command, the specified en_methodis executed every time the error notification
daemon finds a matching entry in the error report. In the example above, the
root user will get e-mail identifying the exact error report entry.
Cluster Planning 45
Download from Www.Somanuals.com. All Manuals Search And Download.
2.6.2.1 Single Point-of-Failure Hardware Component Recovery
HPS Switch network is one resource that has to be considered as a single
point of failure. Since a node can support only one switch adapter, its failure
will disable the switch network for this node. It is strongly recommended to
promote a failure like this into a node failure, if the switch network is critical to
your operations.
Critical failures of the switch adapter would cause an entry in the AIX error
log. Error labels like HPS_FAULT9_ERor HPS_FAULT3_ERare considered critical,
and can be specified to AIX Error Notification in order to be able to act upon
them.
With HACMP, there is a SMIT screen to make it easier to set up an error
notification object. This is much easier than the traditional AIX way of adding
a template file to the ODM class. Under smit hacmp > RAS Support > Error
Notification > Add a Notify Method, you will find the menu allowing you to
add these objects to the ODM. An example of the SMIT panel is shown
below:
Add a Notify Method
Type or select values in entry fields.
Press Enter AFTER making all desired changes.
[Entry Fields]
* Notification Object Name
* Persist across system restart?
Process ID for use by Notify Method
Select Error Class
[HPS_ER9]
Yes
[]
+
+#
+
All
Select Error Type
Match Alertable errors?
Select Error Label
PERM
All
+
+
[HPS_FAULT9_ER] +
Resource Name
Resource Class
Resource Type
[All]
[All]
[All]
+
+
+
* Notify Method
[/usr/sbin/cluster/utilities/clstop -grsy]
F1=Help
F5=Reset
F9=Shell
F2=Refresh
F6=Command
F10=Exit
F3=Cancel
F7=Edit
Enter=Do
F4=List
F8=Image
Figure 8. Sample Screen for Add a Notification Method
46
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The above example screen will add a Notification Method to the ODM, so that
upon appearance of the HPS_FAULT9_ER entry in the error log, the error
notification daemon will trigger the execution of the
/usr/sbin/cluster/utilities/clstop -grsy command, which shuts HACMP down
gracefully with takeover. In this way, the switch failure is acted upon as a
node failure.
2.6.2.2 Notification
The method that is triggered upon the appearance of a specified error log
entry will be run by the error notification daemon with the command sh -c
<en_method>. Because this a regular shell, any shell script can act as a
method.
So, if you want a specific notification, such as e-mail from this event, you can
define a script that sends e-mail and then issues the appropriate commands.
Note
Because the Notification Method is an object in the node’s ODM, it has to
be added to each and every node potentially facing a situation where it
would be wise to act upon the appearance of an error log entry.
This is NOT handled by the HACMP synchronization facility. You have to
take care of this manually.
Alternatively, you can always customize any cluster event to enable a Notify
Command whenever this event is triggered through the SMIT screen for
customizing events.
2.6.2.3 Application Failure
Even application failures can cause an event to happen, if you have
configured this correctly. To do so, you have to find some method to decide
whether an application has failed. This can be as easy as looking for a
specific process, or much more complex, depending on the application. If you
issue an Operator Message through the
errlogger <message>
command, you can act on that as you would on an error notification as
Cluster Planning 47
Download from Www.Somanuals.com. All Manuals Search And Download.
2.7 User ID Planning
The following sections describe various aspects of User ID Planning.
2.7.1 Cluster User and Group IDs
One of the basic tasks any system administrator must perform is setting up
user accounts and groups. All users require accounts to gain access to the
system. Every user account must belong to a group. Groups provide an
additional level of security and allow system administrators to manipulate a
group of users as a single entity.
For users of an HACMP for AIX cluster, system administrators must create
duplicate accounts on each cluster node. The user account information
stored in the /etc/passwd file, and in other files stored in the /etc/security
directory, should be consistent on all cluster nodes. For example, if a cluster
node fails, users should be able to log on to the surviving nodes without
experiencing problems caused by mismatches in the user or group IDs.
System administrators typically keep user accounts synchronized across
cluster nodes by copying the key system account and security files to all
cluster nodes whenever a new account is created or an existing account is
changed.Typically rdistor rcpis used, for that. On RS/6000 SP systems pcp
or supperare widely used. For C-SPOC clusters, the C-SPOC utility simplifies
the cluster-wide synchronization of user accounts by propagating the new
account or changes to an existing account across all cluster nodes
automatically.
The following are some common user and group management tasks, and are
• Listing all user accounts on all cluster nodes
• Adding users to all cluster nodes
• Changing characteristics of a user account on all cluster nodes
• Removing a user account from all cluster nodes.
• Listing all groups on all cluster nodes
• Adding groups to all cluster nodes
• Changing characteristics of a group on all cluster nodes
• Removing a group from all cluster nodes
48
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
2.7.2 Cluster Passwords
While user and group management is very much facilitated with C-SPOC, the
password information still has to be distributed by some other means. If the
system is not configured to use NIS or DCE, the system administrator still has
to distribute the password information, meaning that found in the
/etc/security/password file, to all cluster nodes.
As before, this can be done through rdistor rcp. On RS/6000 SP systems,
there are tools like pcpor supperto distribute information or better files.
2.7.3 User Home Directory Planning
As for user IDs, the system administrator has to ensure that users have their
home directories available and in the same position at all times. That is, they
don’t care whether a takeover has taken place or everything is normal. They
simply want to access their files, wherever they may reside physically, under
the same directory path with the same permissions, as they would on a single
machine.
There are different approaches to that. You could either put them on a shared
volume and handle them within a resource group, or you could use NFS
mounts.
2.7.3.1 Home Directories on Shared Volumes
Within an HACMP cluster, this approach is quite obvious, however, it restricts
you to only one machine where a home directory can be active at any given
time. If you have only one application that the user needs to access, or all of
the applications are running on one machine, where the second node serves
as a standby machine only, this would be sufficient.
2.7.3.2 NFS-Mounted Home Directories
The NFS mounted home directory approach is much more flexible. Because
the directory can be mounted on several machines at the same time, a user
can work with it in several applications on several nodes at the same time.
However, if one cluster node provides NFS service of home directories to
other nodes, in case of a failure of the NFS server node, the access to the
home directories is barred. Placing them onto a machine outside the cluster
doesn’t help either, since this again introduces a single point of failure, and
machines outside the cluster are not any less likely to fail than machines
within.
Cluster Planning 49
Download from Www.Somanuals.com. All Manuals Search And Download.
2.7.3.3 NFS-Mounted Home Directories on Shared Volumes
So, a combined approach is used in most cases. In order to make home
directories a highly available resource, they have to be part of a resource
group and placed on a shared volume. That way, all cluster nodes can access
them in case they need to.
To make the home directories accessible on nodes that currently do not own
the resource where they are physically residing, they have to be NFS
exported from the resource group and imported on all the other nodes in case
any application is running there, needing access to the users files.
In order to make the directory available to users again, when a failover
happens, the takeover node that previously had the directory NFS mounted
from the failed node has to break locks on NFS files, if there are any. Next, it
must unmount the NFS directory, acquire the shared volume (varyon the
shared volume group) and mount the shared file system. Only after that can
users access the application on the takeover node again.
50
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 3. Cluster Hardware and Software Preparation
This chapter covers the steps that are required to prepare the RS/6000
hardware and AIX software for the installation of HACMP and the
configuration of the cluster. This includes configuring adapters for TCP/IP,
setting up shared volume groups, and mirroring and editing AIX configuration
files.
3.1 Cluster Node Setup
The following sections describe important details of cluster node setup.
3.1.1 Adapter Slot Placement
For information regarding proper adapter placement, see the following
documentation:
• PCI Adapter Placement Reference Guide, SA38-0538
• Adapters, Devices, and Cable Information for Micro Channel Bus
Systems, SA38-0533
• Adapters, Devices, and Cable Information for Multiple Bus Systems,
SA38-0516
3.1.2 Rootvg Mirroring
Of all the components used to build a computer system, physical disk devices
are usually the most susceptible to failure. Because of this, disk mirroring is a
frequently used technique for increasing system availability.
File system mirroring and disk mirroring are easily configured using the AIX
Logical Volume Manager. However, conventional file system and disk
mirroring offer no protection against operating system failure or against a
failure of the disk from which the operating system normally boots.
Operating system failure does not always occur instantaneously, as
demonstrated by a system that gradually loses access to operating system
services. This happens as code and data that were previously being
accessed from memory gradually disappear in response to normal paging.
Normally, in an HACMP environment, it is not necessary to think about
mirroring the root volume group, because the node failure facilities of HACMP
can cover for the loss of any of the rootvg physical volumes. However, it is
possible that a customer with business-critical applications will justify
© Copyright IBM Corp. 1999
51
Download from Www.Somanuals.com. All Manuals Search And Download.
mirroring rootvg in order to avoid the impact of the failover time involved in a
node failure. In terms of maximizing availability, this technique is just as valid
for increasing the availability of a cluster as it is for increasing single-system
availability.
The following procedure contains information that will enable you to mirror
the root volume group (rootvg), using the advanced functions of the Logical
Volume Manager (LVM). It contains the steps required to:
• Mirror all the file systems in rootvg.
• Create an additional boot logical volume (blv).
• Modify the bootlist to contain all boot devices.
You may mirror logical volumes in the rootvg in the same way as any AIX
logical volume may be mirrored, either once (two copies), or twice (three
copies). The following procedure is designed for mirroring rootvg to a second
disk only. Upon completion of these steps, your system will remain available if
one of the disks in rootvg fails, and will even automatically boot from an
alternate disk drive, if necessary.
If the dump device is mirrored, you may not be able to capture the dump
image from a crash or the dump image may be corrupted. The design of LVM
prevents mirrored writes of the dump device. Only one of the mirrors will
receive the dump image. Depending on the boot sequence and disk
availability after a crash, the dump will be in one of the following three states:
1. Not available
2. Available and not corrupted
3. Available and corrupted
State (1) will always be a possibility. If the user prefers to prevent the risk of
encountering State (3), then the user must create a non-mirrored logical
volume (that is not hd6) and set the dump device to this non-mirrored logical
volume.
In AIX 4.2.1, two new LVM commands were introduced: mirrorvgand
unmirrorvg. These two commands where introduced to simplify mirroring or
unmirroring of the entire contents of a volume group. The commands will
detect if the entity to be mirrored or unmirrored is rootvg, and will give slightly
different completion messages based on the type of volume group.The
mirrorvg command does the equivalent of Procedure steps (2), (3), and (4).
The mirrorvgcommand takes dump devices and paging devices into account.
If the dump devices are also the paging device, the logical volume will be
52
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
mirrored. If the dump devices are NOT the paging device, that dump logical
volume will not be mirrored.
3.1.2.1 Procedure
The following steps assume the user has rootvg contained on hdisk0 and is
attempting to mirror the rootvg to a new disk: hdisk1.
1. Extend rootvg to hdisk1 by executing the following:
extendvg rootvg hdisk1
2. Disable QUORUM, by executing the following:
chvg -Qn rootvg
3. Mirror the logical volumes that make up the AIX operating system by
executing the following:
mklvcopy hd1 2 hdisk1 # /home file system
mklvcopy hd2 2 hdisk1 # /usr file system
mklvcopy hd3 2 hdisk1 # /tmp file system
mklvcopy hd4 2 hdisk1 # / (root) file system
mklvcopy hd5 2 hdisk1 # blv, boot logical volume
mklvcopy hd6 2 hdisk1 # paging space
mklvcopy hd8 2 hdisk1 # file system log
mklvcopy hd9var 2 hdisk1 # /var file system
If you have other paging devices, rootvg and non-rootvg, it is
recommended that you also mirror those logical volumes in addition to
hd6.
If hd5 consists of more than one logical partition, then, after mirroring hd5
you must verify that the mirrored copy of hd5 resides on contiguous
physical partitions. This can be verified with the following command:
lslv -m hd5
If the mirrored hd5 partitions are not contiguous, you must delete the
mirror copy of hd5 (on hdisk1) and rerun the mklvcopyfor hd5, using the
Cluster Hardware and Software Preparation 53
Download from Www.Somanuals.com. All Manuals Search And Download.
“-m” option. You should consult documentation on the usage of the “-m”
option for mklvcopy.
4. Synchronize the newly created mirrors with the following command:
syncvg -v rootvg
5. Bosboot to initialize all boot records and devices by executing the
following command:
bosboot -a -d /dev/hdisk?
where hdisk?is the first hdisk listed under the “PV” heading after the
command lslv -l hd5 has executed.
6. Initialize the boot list by executing the following:
bootlist -m normal hdisk0 hdisk1
Note
Even though this command identifies the list of possible boot disks, it
does not guarantee that the system will boot from the alternate disk in all
cases involving failures of the first disk. In such situations, it may be
necessary for the user to boot from the installation/maintenance media.
Select maintenance, reissue the bootlistcommand leaving out the
failing disk, and then reboot. On some models, firmware provides a
utility for selecting the boot device at boot time. This may also be used
to force the system to boot from the alternate disk.
7. Shutdown and reboot the system by executing the following command:
shutdown -Fr
This is so that the “Quorum OFF” functionality takes effect.
54
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
3.1.2.2 Necessary APAR Fixes
Table 11. Necessary APAR Fixes
AIX Version
APARs needed
4.1
IX56564
IX61184
IX60521
4.2
IX62417
IX68483
IX70884
IX72058
4.3
IX72550
To determine if either fix is installed on a machine, execute the following:
instfix -i -k <apar number>
3.1.3 AIX Prerequisite LPPs
In order to install HACMP and HACMP/ES the AIX setup must be in a proper
state. The following table gives you the prerequisite AIX levels for the
different HACMP versions:
Table 12. AIX Prerequisite LPPs
HACMP Version
Prerequisite AIX and PSSP Version
HACMP 4.1 for AIX
AIX 4.1.5
PSSP 2.2, if installed on an SP
HACMP 4.2 for AIX
HACMP 4.3 for AIX
HACMP/ES 4.2 for AIX
HACMP/ES 4.3 for AIX
AIX 4.1.5
PSSP 2.2, if installed on an SP
AIX 4.3.2
PSSP 2.2, if installed on an SP
AIX 4.2.1
PSSP 2.2, if installed on an SP
AIX 4.3.2
PSSP 3.1, if installed on an SP
The Prerequisites for the HACMP component HAView 4.2 are
• xlC.rte 3.1.3.0
• nv6000.base.obj 4.1.0.0
Cluster Hardware and Software Preparation 55
Download from Www.Somanuals.com. All Manuals Search And Download.
• nv6000.database.obj 4.1.0.0
• nv6000.Features.obj 4.1.2.0
• nv6000.client.obj 4.1.0.0
and for HAView 4.3
• xlC.rte 3.1.4.0
• nv6000.base.obj 4.1.2.0
• nv6000.database.obj 4.1.2.0
• nv6000.Features.obj 4.1.2.0
• nv6000.client.obj 4.1.2.0
3.1.4 AIX Parameter Settings
This section discusses several general tasks necessary to ensure that your
HACMP for AIX cluster environment works as planned. Consider or check the
following issues to ensure that AIX works as expected in an HACMP cluster.
• I/O pacing
• Network option settings
• /etc/hosts file and nameserver edits
• /.rhosts file edits
3.1.4.1 I/O Pacing
AIX users have occasionally seen poor interactive performance from some
applications when another application on the system is doing heavy
input/output. Under certain conditions, I/O can take several seconds to
complete. While the heavy I/O is occurring, an interactive process can be
severely affected if its I/O is blocked, or, if it needs resources held by a
blocked process.
Under these conditions, the HACMP for AIX software may be unable to send
keepalive packets from the affected node. The Cluster Managers on other
cluster nodes interpret the lack of keepalives as node failure, and the
I/O-bound node is “failed” by the other nodes. When the I/O finishes, the node
resumes sending keepalives. Its packets, however, are now out of sync with
the other nodes, which then kill the I/O-bound node with a RESET packet.
You can use I/O pacing to tune the system so that system resources are
distributed more equitably during high disk I/O. You do this by setting high-
56
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
and low-water marks. If a process tries to write to a file at the high-water
mark, it must wait until enough I/O operations have finished to make the
low-water mark.
Use the smit chgsysfastpath to set high- and low-water marks on the
Change/Show Characteristics of the Operating System screen.
By default, AIX is installed with high- and low-water marks set to zero, which
disables I/O pacing.
While enabling I/O pacing may have a slight performance effect on very
I/O-intensive processes, it is required for an HACMP cluster to behave
correctly during large disk writes. If you anticipate heavy I/O on your HACMP
cluster, you should enable I/O pacing.
Although the most efficient high- and low-water marks vary from system to
system, an initial high-water mark of 33 and a low-water mark of 24 provides
a good starting point. These settings only slightly reduce write times and
consistently generate correct fallover behavior from the HACMP for AIX
software.
See the AIX Performance Monitoring & Tuning Guide, SC23-2365, for more
information on I/O pacing.
3.1.4.2 Checking Network Option Settings
To ensure that HACMP for AIX requests for memory are handled correctly,
you can set (on every cluster node) thewall network option to be higher than
its default value. The suggested value for this option is shown below:
thewall = 5120
To change this default value, add the following line to the end of the
/etc/rc.net file:
no -o thewall=5120
After making this change, monitor mbuf usage using the netstat -m command
and increase or decrease “thewall” option as needed.
To list the values of other network options (not configurable) that are currently
set on a node, enter:
no -a
Cluster Hardware and Software Preparation 57
Download from Www.Somanuals.com. All Manuals Search And Download.
3.1.4.3 Editing the /etc/hosts File and Nameserver Configuration
Make sure all nodes can resolve all cluster addresses. See the chapter on
planning TCP/IP networks (the section Using HACMP with NIS and DNS) in
the HACMP for AIX, Version 4.3: Planning Guide, SC23-4277 for more
information on name serving and HACMP.
Edit the /etc/hosts file (and the /etc/resolv.conf file, if using the nameserver
configuration) on each node in the cluster to make sure the IP addresses of
all clustered interfaces are listed.
For each boot address, make an entry similar to the following:
100.100.50.200 crab_boot
Also, make sure that the /etc/hosts file on each node has the following entry:
127.0.0.1 loopback localhost
3.1.4.4 cron and NIS Considerations
If your HACMP cluster nodes use NIS services, which include the mapping of
the /etc/passwd file, and IPAT is enabled, users that are known only in the
NIS-managed version of the /etc/passwd file will not be able to create
crontabs. This is because cron is started with the /etc/inittab file with run level
2 (for example, when the system is booted), but ypbind is started in the
course of starting HACMP with the rcnfsentry in /etc/inittab. When IPAT is
enabled in HACMP, the run level of the rcnfsentry is changed to -aand run
with the telinit -acommand by HACMP.
In order to let those NIS-managed users create crontabs, you can do one of
the following:
• Change the run level of the cronentry in /etc/inittabto -aand make sure
it is positioned after the rcnfsentry in /etc/inittab. This solution is
recommended if it is acceptable to start cron after HACMP has started.
• Add an entry to the /etc/inittab file like the following script with run level -a.
Make sure it is positioned after the rcnfsentry in /etc/inittab. The important
thing is to kill the cron process, which will respawn and know about all of
the NIS-managed users. Whether or not you log the fact that cron has
been refreshed is optional.
58
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
#! /bin/sh
# This script checks for a ypbind and a cron process. If both
# exist and cron was started before ypbind, cron is killed so
# it will respawn and know about any new users that are found
# in the passwd file managed as an NIS map.
echo "Entering $0 at ‘date‘" >> /tmp/refr_cron.out
cronPid=‘ps -ef |grep "/etc/cron" |grep -v grep |awk \
’{ print $2 }’‘
ypbindPid=‘ps -ef | grep "/usr/etc/ypbind" | grep -v grep | \
if [ ! -z "${ypbindPid}" ]
then
if [ ! -z "${cronPid}" ]
then
echo "ypbind pid is ${ypbindPid}" >> /tmp/refr_cron.out
echo "cron pid is ${cronPid}" >> /tmp/refr_cron.out
echo "Killing cron(pid ${cronPid}) to refresh user \
list" >> /tmp/refr_cron.out
kill -9 ${cronPid}
if [ $? -ne 0 ]
then
echo "$PROGNAME: Unable to refresh cron." \
>>/tmp/refr_cron.out
exit 1
fi
fi
fi
echo "Exiting $0 at ‘date‘" >> /tmp/refr_cron.out
exit 0
3.1.4.5 Editing the /.rhosts File
Make sure that each node’s service adapters and boot addresses are listed in
the /.rhosts file on each cluster node. Doing so allows the
/usr/sbin/cluster/utilities/clruncmdcommand and the
/usr/sbin/cluster/godm daemon to run. The /usr/sbin/cluster/godm daemon is
used when nodes are configured from a central location.
For security reasons, IP label entries that you add to the /.rhosts file to
identify cluster nodes should be deleted when you no longer need to log on to
a remote node from these nodes. The cluster synchronization and verification
functions use rcmdand rshand thus require these /.rhostsentries. These
entries are also required to use C-SPOC commands in a cluster environment.
The /usr/sbin/cluster/clstrmgr daemon, however, does not depend on /.rhosts
file entries.
The /.rhosts file is not required on SP systems running the HACMP Enhanced
Security. This feature removes the requirement of TCP/IP access control lists
(for example, the /.rhosts file) on remote nodes during HACMP configuration.
Cluster Hardware and Software Preparation 59
Download from Www.Somanuals.com. All Manuals Search And Download.
3.2 Network Connection and Testing
The following sections describe important aspects of network connection and
testing.
3.2.1 TCP/IP Networks
Since there are several types of TCP/IP Networks available within HACMP,
there are several different characteristics and some restrictions on them.
Characteristics like maximum distance between nodes have to be
considered. You don’t want to put two cluster nodes running a mission-critical
application in the same room for example.
3.2.1.1 Cabling Considerations
Characteristics of the different types of cable, their maximum length, and the
like are beyond the scope of this book. However, for actual planning of your
clusters, you have to check whether your network cabling allows you to put
two cluster nodes away from each other, or even in different buildings.
There’s one additional point with cabling, that should be taken care of.
Cabling of networks often involves hubs or switches. If not carefully planned,
this sometimes introduces another single point of failure into your cluster. To
eliminate this you should have at least two hubs.
As shown in Figure 9, failure of a hub would not result in one machine being
disconnected from the network. In that case, a hub failure would cause either
both service adapters to fail, which would cause a swap_adapter event, and
the standby adapters would take over the network, or both standby adapters
would fail, which would cause fail_standby events. Configuring a notify
method for these events can alert the network administrator to check and fix
the broken hub.
60
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
.
Figure 9. Connecting Networks to a Hub
3.2.1.2 IP Addresses and Subnets
The design of the HACMP for AIX software specifies that:
• All client traffic be carried over the service adapter
• Standby adapters be hidden from client applications and carry only
internal Cluster Manager traffic
Cluster Hardware and Software Preparation 61
Download from Www.Somanuals.com. All Manuals Search And Download.
To comply with these rules, pay careful attention to the IP addresses you
assign to standby adapters. Standby adapters must be on a separate subnet
from the service adapters, even though they are on the same physical
network. Placing standby adapters on a different subnet from the service
adapter allows HACMP for AIX to determine which adapter TCP/IP will use to
send a packet to a network.
If there is more than one adapter with the same network address, there is no
way to guarantee which of these adapters will be chosen by IP as the
transmission route. All choices will be correct, since each choice will deliver
the packet to the correct network. To guarantee that only the service adapter
handles critical traffic, you must limit IP’s choice of a transmission route to
one adapter. This keeps all traffic off the standby adapter so that it is
available for adapter swapping and IP address takeover (IPAT). Limiting the
IP’s choice of a transmission route also facilitates identifying an adapter
failure.
Note
The netmask for all adapters in an HACMP network must be the same even
though the service and standby adapters are on different logical subnets.
See the HACMP for AIX, Version 4.3: Concepts and Facilities, SC23-4276
guide for more information about using the same netmask for all adapters.
information.
3.2.1.3 Testing
After setting up all adapters with AIX, you can do several things to check
whether TCP/IP is working correctly. Note, that without HACMP being started,
the service adapters defined to HACMP will remain on their boot address.
After startup these adapters change to their service addresses.
Use the following AIX commands to investigate the TCP/IP subsystem:
• Use the netstatcommand to make sure that the adapters are initialized
and that a communication path exists between the local node and the
target node.
• Use the pingcommand to check the point-to-point connectivity between
nodes.
• Use the ifconfigcommand on all adapters to detect bad IP addresses,
incorrect subnet masks, and improper broadcast addresses.
62
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Scan the /tmp/hacmp.out file to confirm that the /etc/rc.net script has run
successfully. Look for a zero exit status.
• If IP address takeover is enabled, confirm that the /etc/rc.net script has
run and that the service adapter is on its service address and not on its
boot address.
• Use the lssrc -g tcpipcommand to make sure that the inetd daemon is
running.
• Use the lssrc -g portmapcommand to make sure that the portmapper
daemon is running.
• Use the arpcommand to make sure that the cluster nodes are not using
the same IP or hardware address.
3.2.2 Non TCP/IP Networks
Currently three types of non-TCP/IP networks are supported:
• Serial (RS232)
• Target-mode SCSI
• Target-mode SSA
While we use the word serial here to refer to RS232 only, in HACMP
definitions, a “serial” network means a non-TCP/IP network of any kind.
Therefore, when we are talking about HACMP network definitions, a serial
network could also be a target-mode SCSI or target-mode SSA network.
The following describes some cabling issues on each type of non-TCP/IP
network, how they are to be configured, and how you can test if they are
operational.
3.2.2.1 Cabling Considerations
RS232
Cabling a serial connection requires a null-modem cable. As often
cluster nodes are further apart than 60 m (181 ft.), sometimes
modem eliminators or converters to fiber channel are used.
TMSCSI If your cluster uses SCSI disks as shared devices, you can use
that line for TMSCSI as well. TMSCSI requires Differential SCSI
page 26). Because the SCSI bus has to be terminated on both
ends, and not anywhere else in between, resistors on the
adapters should be removed, and cabling should be done as
shown in Figure 11 on page 77, that is, with Y-cables that are
terminated at one end connected to the adapters where the other
end connects to the shared disk device.
Cluster Hardware and Software Preparation 63
Download from Www.Somanuals.com. All Manuals Search And Download.
TMSSA
Target-mode SSA is only supported with the SSA Multi-Initiator
RAID Adapters (Feature #6215 and #6219), Microcode Level
1801 or later. You need at least HACMP Version 4.2.2 with APAR
IX75718.
3.2.2.2 Configuring RS232
Use the smit ttyfastpath to create a tty device on the nodes. On the resulting
panel, you can add an RS232 tty by selecting a native serial port, or a port on
an asynchronous adapter. Make sure that the Enable Login field is set to
disable. You do not want a getty process being spawned on this interface.
3.2.2.3 Configuring Target Mode SCSI
To configure a target-mode SCSI network on the Differential SCSI adapters,
you have to enable the SCSI adapter’s feature TARGET MODE by setting the
enabled characteristics to yes. Since disks on the SCSI bus are normally
configured at boot time, and the characteristics of the parent device cannot
be changed as long as there are child devices present and active, you have
to set all the disks on that bus to Definedwith the
rmdev -l hdiskx
command, before you can enable that feature. Alternatively you can make
these changes to the database (ODM) only, and they will be activated at the
time of the next reboot.
If you choose not to reboot, instead setting all the child devices to Defined,
you have to run cfgmgr, to get the tmscsi device created, as well as all the
child devices of the adapter back to the available state.
Note
The target mode device created is a logical new device on the bus.
Because it is created by scanning the bus for possible initiator devices, a
tmscsix device is created on a node for each SCSI adapter on the same
bus that has the target mode flag enabled, therefore representing this
adapter’s unique SCSI ID. In that way, the initiator can address packets to
exactly one target device.
This procedure has to be done for all the cluster nodes that are going to use a
serial network of type tmscsi as defined in your planning sheets.
64
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
3.2.2.4 Configuring Target Mode SSA
The node number on each system needs to be changed from the default of
zero to a number. All systems on the SSA loop must have a unique node
number.
To change the node number use the following command:
chdev -l ssar -a node_number=#
To show the system’s node number use the following command:
lsattr -El ssar
Having the node numbers set to non-zero values enables the target mode
devices to be configured. Run the cfgmgrcommand to configure the tmssa#
devices on each system. Check that the tmssa devices are available on each
system using the following command:
lsdev -C | grep tmssa
The Target Mode SCSI or SSA serial network can now be configured into an
HACMP cluster.
3.2.2.5 Testing RS232 and Target Mode Networks
Testing of the serial networks functionality is similar. Basically you just write
to one side’s device and read from the other.
Serial (RS323): After configuring the serial adapter and cabling it correctly,
you can check the functionality of the connection by entering the command
cat < /dev/ttyx
on one node for reading from that device and
cat /etc/environment > /dev/ttyy
on the corresponding node for writing. You should see the first command
hanging until the second command is issued, and then showing the output of
it.
Target Mode SSA: After configuration of Target Mode SSA, you can check
the functionality of the connection by entering the command:
cat < /dev/tmssax.tm
on one node for reading from that device and:
Cluster Hardware and Software Preparation 65
Download from Www.Somanuals.com. All Manuals Search And Download.
cat /etc/environment > /dev/tmssay.im
on the corresponding node for writing. x and y correspond to the appropriate
opposite nodenumber. You should see the first command hanging until the
second command is issued, and then showing its output.
Target Mode SCSI: After configuration of Target Mode SCSI, you can check
the functionality of the connection by entering the command:
cat < /dev/tmscsix.tm
on one node for reading from that device and:
cat /etc/environment > /dev/tmscsiy.im
on the corresponding node for writing. You should see the first command
hanging until the second command is issued, and then showing the output of
that second command.
3.3 Cluster Disk Setup
The following sections relate important information about cluster disk setup.
3.3.1 SSA
The following sections describe cabling, AIX configuration, microcode
loading, and configuring a RAID on SSA disks.
3.3.1.1 Cabling
The following rules must be followed when connecting a 7133 SSA
Subsystem:
• Each SSA loop must be connected to a valid pair of connectors on the
SSA adapter card (A1 and A2 to form one loop, or B1 and B2 to form one
loop).
• Only one pair of connectors of an SSA adapter can be connected in a
particular SSA loop (A1 or A2, with B1 or B2 cannot be in the same SSA
loop).
• A maximum of 48 disks can be connected in an SSA loop.
• A maximum of three dummy disk drive modules can be connected next to
each other.
• The maximum length of an SSA cable is 25 m. With Fiber-Optic
Extenders, the connection length can be up to 2.4 km.
66
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
For more information regarding adapters and cabling rules see 2.3.1, “SSA
Disks” on page 16 or the following documents:
• 7133 SSA Disk Subsystems: Service Guide, SY33-0185-02
• 7133 SSA Disk Subsystem: Operator Guide, GA33-3259-01
• 7133 Models 010 and 020 SSA Disk Subsystems: Installation Guide,
GA33-3260-02
• 7133 Models 500 and 600 SSA Disk Subsystems: Installation Guide,
GA33-3263-02
• 7133 SSA Disk Subsystems for Open Attachment: Service Guide,
SY33-0191-00
• 7133 SSA Disk Subsystems for Open Attachment: Installation and User's
Guide, SA33-3273-00
3.3.1.2 AIX Configuration
During boot time, the configuration manager of AIX configures all the device
drivers needed to have the SSA disks available for usage. The configuration
manager can’t do this configuration if the SSA Subsystem is not properly
connected or if the SSA Software is not installed. If the SSA Software is not
already installed, the configuration manager will tell you the missing filesets.
You can either install the missing filesets with smit, or call the configuration
manager with the -i flag.
The configuration manager configures the following devices:
• SSA Adapter Router
• SSA Adapter
• SSA Disks
Adapter Router
The adapter Router (ssar) is only a conceptual configuration aid and is
always in a “Defined” state. It cannot be made “Available.” You can list the
ssar with the following command:
#lsdev -C | grep ssar
ssar
Defined
SSA Adapter Router
Cluster Hardware and Software Preparation 67
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter Definitions
By issuing the following command, you can check the correct adapter
configuration. In order to work correctly, the adapter must be in the
“Available” state:
#lsdev -C | grep ssa
ssa0
ssar
Available 00-07
Defined
SSA Enhanced Adapter
SSA Adapter Router
The third column in the adapter device line shows the location of the adapter.
Disk Definitions
SSA disk drives are represented in AIX as SSA logical disks (hdisk0,
hdisk1,...,hdiskN) and SSA physical disks (pdisk0, pdisk1,...,pdiskN). SSA
RAID arrays are represented as SSA logical disks (hdisk0, hdisk1,...,hdiskN).
SSA logical disks represent the logical properties of the disk drive or array,
and can have volume groups and file systems mounted on them. SSA
physical disks represent the physical properties of the disk drive. By default,
one pdisk is always configured for each physical disk drive. One hdisk is
configured for each disk drive that is connected to the using system, or for
each array. By default, all disk drives are configured as system (AIX) disk
drives. The array management software can be used to change the disks
from hdisks to array candidate disks or hot spares.
SSA logical disks:
• Are configured as hdisk0, hdisk1,...,hdiskN.
• Support a character special file (/dev/rhdisk0,
/dev/rhdisk1,...,/dev/rhdiskN).
• Support a block special file (/dev/hdisk0, /dev/hdisk1,...,/dev/hdiskN).
• Support the I/O Control (IOCTL) subroutine call for non service and
diagnostic functions only.
• Accept the read and write subroutine calls to the special files.
• Can be members of volume groups and have file systems mounted on
them.
In order to list the logical disk definitions, use the following command:
68
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
#lsdev -Cc disk| grep SSA
hdisk3 Available 00-07-L
hdisk4 Available 00-07-L
hdisk5 Available 00-07-L
hdisk6 Available 00-07-L
hdisk7 Available 00-07-L
hdisk8 Available 00-07-L
SSA Logical Disk Drive
SSA Logical Disk Drive
SSA Logical Disk Drive
SSA Logical Disk Drive
SSA Logical Disk Drive
SSA Logical Disk Drive
SSA physical disks:
• Are configured as pdisk0, pdisk1,...,pdiskN.
• Have errors logged against them in the system error log.
• Support a character special file (/dev/pdisk0, /dev/pdisk1,...,/dev/p.diskN).
• Support the IOCTLl subroutine for servicing and diagnostic functions.
• Do not accept read or write subroutine calls for the character special file.
In order to list the physical disk definitions use the following command:
#lsdev -Cc pdisk| grep SSA
pdisk0 Available 00-07-P 1GB SSA C Physical Disk Drive
pdisk1 Available 00-07-P 1GB SSA C Physical Disk Drive
pdisk2 Available 00-07-P 1GB SSA C Physical Disk Drive
pdisk3 Available 00-07-P 1GB SSA C Physical Disk Drive
pdisk4 Available 00-07-P 1GB SSA C Physical Disk Drive
pdisk5 Available 00-07-P 1GB SSA C Physical Disk Drive
Diagnostics
A good tool to get rid of SSA problems are the SSA service aids in the AIX
diagnostic program diag.The SSA diagnostic routines are fully documented
in A Practical Guide to SSA for AIX, SG24-4599. The following is a brief
overview:
The SSA service aids are accessed from the main menu of the diag program.
Select Task Selection -> SSA Service Aids. This will give you the following
options:
Set Service Mode
This option enables you to determine the location
of a specific SSA disk drive within a loop and to
remove the drive from the configuration, if
required.
Link Verification
This option enables you to determine the
operational status of a link
Cluster Hardware and Software Preparation 69
Download from Www.Somanuals.com. All Manuals Search And Download.
Configuration Verification This option enables you to display the
relationships between physical (pdisk) and logical
(hdisk) disks.
Format Disk
Certify Disk
This option enables you to format SSA disk drives.
This option enables you to test whether data on an
SSA disk drive can be read correctly.
Display/Download...
This option enables you to display the microcode
level of the SSA disk drives and to download new
microcode to individual or all SSA disk drives
connected to the system.
Note
When an SSA loop is attached to multiple host systems, do not invoke the
diagnostic routines from more than one host simultaneously, to avoid
unpredictable results that may result in data corruption.
3.3.1.3 Microcode Loading
To ensure that everything works correctly, install the latest filesets, fixes and
microcode for your SSA disk subsystem. The latest information and
downloadable files can be found under http://www.hursley.ibm.com/~ssa.
Upgrade Instructions
Follow these steps to perform an upgrade:
1. Login as root
2. Download the appropriate microcode file for your AIX version from the
web-site mentioned above
3. Save the file upgrade.tar in your /tmp directory
4. Type tar -xvf upgrade.tar
5. Run smitty install
6. Select install & update software
7. Select install & update from ALL available software
8. Use the directory /usr/sys/inst.images as the install device
9. Select all filesets in this directory for install
10.Execute the command
11.Exit Smit
70
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Note
You must ensure that:
• You do not attempt to perform this adapter microcode download
concurrently on systems that are in the same SSA loop. This may cause
a portion of the loop to be isolated and could prevent access to these
disks from elsewhere in the loop.
• You do not run advanced diagnostics while downloads are in progress.
Advanced diagnostics causes the SSA adapter to be reset temporarily,
thereby introducing a break in the loop, portions of the loop may
become temporarily isolated and inaccessible.
• You have complete SSA loops. Check this by using diagnostics in
System Verification mode. If you have incomplete loops (such as
strings) action must be taken to resolve this before you can continue.
• All of your loops are valid, in this case with one or two adapters in each
loop. This is also done by using Diagnostics in System Verification
mode.
12.Run cfgmgr to install the microcode to adapters.
13.To complete the device driver upgrade, you must now reboot your system.
14.To confirm that the upgrade was a success, type lscfg -vl ssaXwhere X is
0,1... for all SSA adapters. Check the ROS Level line to see that each
adapter has the appropriate microcode level (for the correct microcode
level, see the above mentioned web-site).
15.Run lslpp -l|grep SSAand check that the fileset levels you have match, or
are above the levels shown in the list on the above mentioned web-site. If
any of the SSA filesets are at a lower level than those shown in the above
link, please repeat the whole upgrade procedure again. If, after repeating
the procedure, the code levels do not match the latest ones, place a call
with your local IBM Service Center.
16.If the adapters are in SSA loops which contain other adapters in other
systems, please repeat this procedure on all systems as soon as possible.
17.In order to install the disk microcode, run ssadload -ufrom each system in
turn.
Note
Allow ssadload to complete on one system before running it on another.
Cluster Hardware and Software Preparation 71
Download from Www.Somanuals.com. All Manuals Search And Download.
18.To confirm that the upgrade was a success, type lscfg -vl pdiskXwhere X
is 0,1... for all SSA disks. Check the ROS Level line to see that each disk
has the appropriate microcode level (for the correct microcode level see
the above mentioned web-site).
3.3.1.4 Configuring a RAID on SSA Disks
Disk arrays are groups of disk drives that act like one disk as far as the
operating system is concerned, and which provide better availability or
performance characteristics than the individual drives operating alone.
Depending on the particular type of array that is used, it is possible to
optimize availability or performance, or to select a compromise between both.
The SSA Enhanced Raid adapters only support RAID level 5 (RAID5). RAID0
(Striping) and RAID1 (Mirroring) is not directly supported by the SSA
Enhanced Raid adapters, but with the Logical Volume Manager (LVM), RAID0
and RAID1 can be configured on non-RAID disks.
In order to create a RAID5 on SSA Disks, use the command smitty ssaraid.
This will show you the following menu:
SSA RAID Arrays
Move cursor to desired item and press Enter.
List All Defined SSA RAID Arrays
List All Supported SSA RAID Arrays
List All SSA RAID Arrays Connected to a RAID Manager
List Status Of All Defined SSA RAID Arrays
List/Identify SSA Physical Disks
List/Delete Old RAID Arrays Recorded in an SSA RAID Manager
Add an SSA RAID Array
Delete an SSA RAID Array
Change/Show Attributes of an SSA RAID Array
Change Member Disks in an SSA RAID Array
Change/Show Use of an SSA Physical Disk
Change Use of Multiple SSA Physical Disks
F1=Help
F9=Shell
F2=Refresh
F10=Exit
F3=Cancel
Enter=Do
F8=Image
Select Add an SSA RAID Array to do the definitions.
3.3.2 SCSI
The following sections contain important information about SCSI: cabling,
connecting RAID subsystems, and adapter SCSI ID and termination change.
72
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
3.3.2.1 Cabling
The following sections describe important information about cabling.
SCSI Adapters
A overview of SCSI adapters that can be used on a shared SCSI bus is given
RAID Enclosures
The 7135 RAIDiant Array can hold a maximum of 30 single-ended disks in
two units (one base and one expansion). It has one controller by default, and
another controller can be added for improved performance and availability.
Each controller takes up one SCSI ID. The disks sit on internal single-ended
buses and hence do not take up IDs on the external bus. In an HACMP
cluster, each 7135 should have two controllers, each of which is connected to
a separate shared SCSI bus. This configuration protects you against any
failure (SCSI adapter, cables, or RAID controller) on either SCSI bus.
Because of cable length restrictions, a maximum of two 7135s on a shared
SCSI bus are supported by HACMP.
3.3.2.2 Connecting RAID Subsystems
In this section, we will list the different components required to connect RAID
subsystems on a shared bus. We will also show you how to connect these
components together.
The 7135-110 RAIDiant Array can be connected to multiple systems on either
an 8-bit or a 16-bit SCSI-2 differential bus. The Model 210 can only be
connected to a 16-bit SCSI-2 Fast/Wide differential bus, using the Enhanced
SCSI-2 Differential Fast/Wide Adapter/A.
To connect a set of 7135-110s to SCSI-2 Differential Controllers on a shared
8-bit SCSI bus, you need the following:
• SCSI-2 Differential Y-Cable
FC: 2422 (0.765m), PN: 52G7348
• SCSI-2 Differential System-to-System Cable
FC: 2423 (2.5m), PN: 52G7349
This cable is used only if there are more than two nodes attached to the
same shared bus.
• Differential SCSI Cable (RAID Cable)
FC: 2901 or 9201 (0.6m), PN: 67G1259 - OR -
Cluster Hardware and Software Preparation 73
Download from Www.Somanuals.com. All Manuals Search And Download.
FC: 2902 or 9202 (2.4m), PN: 67G1260 - OR -
FC: 2905 or 9205 (4.5m), PN: 67G1261 - OR -
FC: 2912 or 9212 (12m), PN: 67G1262 - OR -
FC: 2914 or 9214 (14m), PN: 67G1263 - OR -
FC: 2918 or 9218 (18m), PN: 67G1264
• Terminator (T)
Included in FC 2422 (Y-Cable), PN: 52G7350
• Cable Interposer (I)
FC: 2919, PN: 61G8323
One of these is required for each connection between an SCSI-2
Differential Y-Cable and a Differential SCSI Cable going to the 7135
unit, as shown in Figure 10.
Figure 10 shows four RS/6000s, each represented by two SCSI-2 Differential
Controllers, connected on two 8-bit buses to two 7135-110s, each with two
controllers.
Figure 10. 7135-110 RAIDiant Arrays Connected on Two Shared 8-Bit SCSI Buses
To connect a set of 7135s to SCSI-2 Differential Fast/Wide Adapter/As or
Enhanced SCSI-2 Differential Fast/Wide Adapter/As on a shared 16-bit SCSI
bus, you need the following:
• 16-Bit SCSI-2 Differential Y-Cable
74
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
FC: 2426 (0.94m), PN: 52G4234
• 16-Bit SCSI-2 Differential System-to-System Cable
FC: 2424 (0.6m), PN: 52G4291 - OR -
FC: 2425 (2.5m), PN: 52G4233
This cable is used only if there are more than two nodes attached to the
same shared bus.
• 16-Bit Differential SCSI Cable (RAID Cable)
FC: 2901 or 9201 (0.6m), PN: 67G1259 - OR -
FC: 2902 or 9202 (2.4m), PN: 67G1260 - OR -
FC: 2905 or 9205 (4.5m), PN: 67G1261 - OR -
FC: 2912 or 9212 (12m), PN: 67G1262 - OR -
FC: 2914 or 9214 (14m), PN: 67G1263 - OR -
FC: 2918 or 9218 (18m), PN: 67G1264
• 16-Bit Terminator (T)
Included in FC 2426 (Y-Cable), PN: 61G8324
Figure 11 shows four RS/6000s, each represented by two SCSI-2 Differential
Fast/Wide Adapter/As connected on two 16-bit buses to two 7135-110s, each
with two controllers.
The 7135-210 requires the Enhanced SCSI-2 Differential Fast/Wide
Adapter/A adapter for connection. Other than that, the cabling is exactly the
same as shown in Figure 11, if you just substitute the Enhanced SCSI-2
Differential Fast/Wide Adapter/A (FC: 2412) for the SCSI-2 Differential
Fast/Wide Adapter/A (FC: 2416) in the picture.
Cluster Hardware and Software Preparation 75
Download from Www.Somanuals.com. All Manuals Search And Download.
#2416 (16-
6 (16-bit)
#2424
#2426
#2426
#2416 (16-b
6-bit)
T
T
#2416 (16-bit)
6 (16-bit )
T
T
Maximum total cable length: 25m
76
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
7135-110
7135-110
#2902
RAID
cable
Controller 1
Controller 1
#2902
#2902
#2902
#2426
#2416 (16-bit)
#2902
RAID
cable
#2426
Controller 2
Controller 2
#2902
#2416 (16-bit)
#2416 (16-bit)
#2416 (16-bit)
#2424
#2426
#2426
#2416 (16-bit)
#2416 (16-bit)
#2416 (16-bit )
T
T
#2416 (16-bit)
T
T
Maximum total cable length: 25m
Figure 11. 7135-110 RAIDiant Arrays Connected on Two Shared 16-Bit SCSI Buses
3.3.2.3 Adapter SCSI ID and Termination change
The SCSI-2 Differential Controller is used to connect to 8-bit disk devices on
a shared bus. The SCSI-2 Differential Fast/Wide Adapter/A or Enhanced
SCSI-2 Differential Fast/Wide Adapter/A is usually used to connect to 16-bit
devices but can also be used with 8-bit devices.
In a dual head-of-chain configuration of shared disks, there should be no
termination anywhere on the bus except at the extremities. Therefore, you
should remove the termination resistor blocks from the SCSI-2 Differential
Controller and the SCSI-2 Differential Fast/Wide Adapter/A or Enhanced
SCSI-2 Differential Fast/Wide Adapter/A. The positions of these blocks (U8
and U26 on the SCSI-2 Differential Controller, and RN1, RN2 and RN3 on the
Cluster Hardware and Software Preparation 77
Download from Www.Somanuals.com. All Manuals Search And Download.
SCSI-2 Differential Fast/Wide Adapter/A and Enhanced SCSI-2 Differential
Termination
Resistor
Blocks
P/N 43G0176
4-2
Figure 12. Termination on the SCSI-2 Differential Controller
Termination
Resistor
Internal 16-bit SE Internal 8-bit SE Blocks
P/N 56G7315
4-6
Figure 13. Termination on the SCSI-2 Differential Fast/Wide Adapters
78
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The ID of an SCSI adapter, by default, is 7. Since each device on an SCSI
bus must have a unique ID, the ID of at least one of the adapters on a shared
SCSI bus has to be changed.
The procedure to change the ID of an SCSI-2 Differential Controller is:
1. At the command prompt, enter smit chgscsi.
2. Select the adapter whose ID you want to change from the list presented to
you.
SCSI Adapter
Move cursor to desired item and press Enter.
scsi0 Available 00-02 SCSI I/O Controller
scsi1 Available 06-02 SCSI I/O Controller
scsi2 Available 08-02 SCSI I/O Controller
scsi3 Available 07-02 SCSI I/O Controller
F1=Help
F8=Image
/=Find
F2=Refresh
F10=Exit
n=Find Next
F3=Cancel
Enter=Do
3. Enter the new ID (any integer from 0 to 7) for this adapter in the Adapter
card SCSI ID field. Since the device with the highest SCSI ID on a bus
gets control of the bus, set the adapter’s ID to the highest available ID. Set
the Apply change to DATABASE only field to yes.
Cluster Hardware and Software Preparation 79
Download from Www.Somanuals.com. All Manuals Search And Download.
Change / Show Characteristics of a SCSI Adapter
Type or select values in entry fields.
Press Enter AFTER making all desired changes.
[Entry Fields]
SCSI Adapter
scsi0
Description
Status
Location
SCSI I/O Controller
Available
00-08
Adapter card SCSI ID
BATTERY backed adapter
DMA bus memory LENGTH
Enable TARGET MODE interface
Target Mode interface enabled
PERCENTAGE of bus memory DMA area
for target mode
[6]
no
[0x202000]
no
no
+#
+
+
+
[50]
+#
Name of adapter code download file
Apply change to DATABASE only
/etc/microcode/8d>
yes
+
F1=Help
F5=Reset
F9=Shell
F2=Refresh
F6=Command
F10=Exit
F3=Cancel
F7=Edit
Enter=Do
F4=List
F8=Image
4. Reboot the machine to bring the change into effect.
The same task can be executed from the command line by entering:
# chdev -l scsi1 -a id=6 -P
Also with this method, a reboot is required to bring the change into effect.
The procedure to change the ID of an SCSI-2 Differential Fast/Wide
Adapter/A or Enhanced SCSI-2 Differential Fast/Wide Adapter/A is almost the
same as the one described above. Here, the adapter that you choose from
the list you get after executing the smit chgsyscommand should be an ascsi
device. Also, as shown below, you need to change the external SCSI ID only.
80
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Change/Show Characteristics of a SCSI Adapter
SCSI adapter
ascsi1
Description
Wide SCSI I/O Control>
Status
Location
Internal SCSI ID
External SCSI ID
WIDE bus enabled
...
Available
00-06
7
[6]
yes
+#
+#
+
Apply change to DATABASE only
yes
The command line version of this is:
# chdev -l ascsi1 -a id=6 -P
As in the case of the SCSI-2 Differential Controller, a system reboot is
required to bring the change into effect.
The maximum length of the bus, including any internal cabling in disk
subsystems, is limited to 19 meters for buses connected to the SCSI-2
Differential Controller, and 25 meters for those connected to the SCSI-2
Differential Fast/Wide Adapter/A or Enhanced SCSI-2 Differential Fast/Wide
Adapter/A.
3.4 Shared LVM Component Configuration
This section describes how to define the LVM components shared by cluster
nodes in an HACMP for AIX cluster environment.
Creating the volume groups, logical volumes, and file systems shared by the
nodes in an HACMP cluster requires that you perform steps on all nodes in
the cluster. In general, you define the components on one node (referred to in
the text as the source node) and then import the volume group on the other
nodes in the cluster (referred to as destination nodes). This ensures that the
ODM definitions of the shared components are the same on all nodes in the
cluster.
Non-concurrent access environments typically use journaled file systems to
manage data, while concurrent access environments use raw logical
volumes. This chapter provides different instructions for defining shared LVM
components in non-concurrent access and concurrent access environments.
Cluster Hardware and Software Preparation 81
Download from Www.Somanuals.com. All Manuals Search And Download.
3.4.1 Creating Shared VGs
The following sections contain information about creating non-concurrent
VGs and VGs for concurrent access.
3.4.1.1 Creating Non-Concurrent VGs
This section covers how to create a shared volume group on the source node
using the SMIT interface. Use the smit mkvgfastpath to create a shared
volume group. Use the default field values unless your site has other
requirements, or unless you are specifically instructed otherwise here.
Table 13. smit mkvg Options (Non-Concurrent)
Options
Description
VOLUME GROUP name
The name of the shared volume group should be
unique within the cluster.
Activate
volume
group Set to no so that the volume group can be
AUTOMATICALLY at system activated as appropriate by the cluster event
restart? scripts.
ACTIVATE volume group after it Set to yes.
is created?
Volume
NUMBER
Group
MAJOR If you are not using NFS, you can use the default
(which is the next available number in the valid
range). If you are using NFS, you must make sure
to use the same major number on all nodes. Use
the lvlstmajor command on each node to
determine a free major number common to all
nodes.
3.4.1.2 Creating VGs for Concurrent Access
The procedure used to create a concurrent access volume group varies
depending on which type of device you are using: serial disk subsystem
(7133) or RAID disk subsystem (7135).
Note
If you are creating (or plan to create) concurrent volume groups on SSA
devices, be sure to assign unique non-zero node numbers through the
SSAR on each cluster node. If you plan to specify SSA disk fencing in your
concurrent resource group, the node numbers are assigned when you
synchronize resources. If you do not specify SSA disk fencing, assign node
numbers using the following command: chdev -l ssar -a node_number=x,
where x is the number to assign to that node. You must reboot the system
to effect the change.
82
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Creating a Concurrent Access Volume Group on Serial Disk
Subsystems
To use a concurrent access volume group, defined on a serial disk subsystem
such as an IBM 7133 disk subsystem, you must create it as a
concurrent-capable volume group. A concurrent-capable volume group can
be activated (varied on) in either non-concurrent mode or concurrent access
mode. To define logical volumes on a concurrent-capable volume group, it
must be varied on in non-concurrent mode.
You can use smit mkvgwith the following options to build the volume group:
Table 14. smit mkvg Options (Concurrent, Non-RAID)
Options
Description
VOLUME GROUP name
Specify name of volume group.
Physical partition SIZE in Accept the default.
megabytes
PHYSICAL VOLUME NAMES
Specify the names of the physical volumes you want
included in the volume group.
Activate
volume
group Set this field to no so that the volume group can be
AUTOMATICALLY at system activated as appropriate by the cluster event scripts.
restart?
ACTIVATE volume group after it Set this field to no.
is created?
Volume
Group
MAJOR Accept the default.
NUMBER
Create VG concurrent capable? Set this field to yes so that the volume group can be
activated in concurrent access mode by the HACMP
for AIX event scripts.
Auto-varyon concurrent mode?
Set this field to no so that the volume group can be
activated as appropriate by the cluster event scripts.
Creating a Concurrent Access Volume Group on RAID Disk
Subsystems
To create a concurrent access volume group on a RAID disk subsystem, such
as an IBM 7135 disk subsystem, follow the same procedure as you would to
create a non-concurrent access volume group. A concurrent access volume
group can be activated (varied on) in either non-concurrent mode or
concurrent access mode. To define logical volumes on a concurrent access
volume group, it must be varied on in non-concurrent mode.
Cluster Hardware and Software Preparation 83
Download from Www.Somanuals.com. All Manuals Search And Download.
Use the smit mkvg fastpath to create a shared volume group. Use the default
field values unless your site has other requirements, or unless you are
specifically instructed otherwise.
Table 15. smit mkvg Options (Concurrent, RAID)
Options
Description
VOLUME GROUP name
The name of the shared volume group should be
unique within the cluster.
Activate
volume
group Set to no so that the volume group can be
AUTOMATICALLY at system activated as appropriate by the cluster event
restart? scripts.
ACTIVATE volume group after it Set to yes.
is created?
Volume
NUMBER
Group
MAJOR While it is only really required when you are using
NFS, it is always good practice in an HACMP
cluster to have a shared volume group have the
same major number on all the nodes that serve it.
Use the lvlstmajor command on each node to
determine a free major number common to all
nodes.
Create VG concurrent capable?
Set this field to no.
3.4.2 Creating Shared LVs and File Systems
Use the smit crjfsfast path to create the shared file system on the source
node. When you create a journaled file system, AIX creates the
corresponding logical volume. Therefore, you do not need to define a logical
volume. You do, however, need to later rename both the logical volume and
the log logical volume for the file system and volume group.
Table 16. smit crjfs Options
Options
Description
Mount AUTOMATICALLY at Make sure this field is set to no.
system restart?
Start Disk Accounting
Make sure this field is set to no.
Renaming a jfslog and Logical Volumes on the Source Node
AIX assigns a logical volume name to each logical volume it creates.
Examples of logical volume names are /dev/lv00 and /dev/lv01. Within an
HACMP cluster, the name of any shared logical volume must be unique. Also,
84
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
the journaled file system log (jfslog) is a logical volume that requires a unique
name in the cluster.
To make sure that logical volumes have unique names, rename the logical
volume associated with the file system and the corresponding jfslog logical
volume. Use a naming scheme that indicates the logical volume is associated
with a certain file system. For example, lvsharefs could name a logical
volume for the /sharefs file system.
1. Use the lsvg -l volume_group_namecommand to determine the name of the
logical volume and the log logical volume (jfslog) associated with the
shared volume groups. In the resulting display, look for the logical volume
name that has type jfs. This is the logical volume. Then look for the logical
volume name that has type jfslog. This is the log logical volume.
2. Use the smit chlvfastpath to rename the logical volume and the log
logical volume.
3. After renaming the jfslog or a logical volume, check the /etc/filesystems
file to make sure the dev and log attributes reflect the change. Check the
log attribute for each file system in the volume group, and make sure that
it has the new jfslog name. Check the dev attribute for the logical volume
that you renamed, and make sure that it has the new logical volume name.
Adding Copies to Logical Volume on the Source Node
Note
These steps do not apply to RAID devices, which provide their own
mirroring of logical volumes.
1. Use the smit mklvcopy fastpath to add copies to a logical volume. Add
copies to both the jfslog log logical volume and the logical volumes in the
shared file systems. To avoid space problems, first mirror the jfslog log
logical volume and then the shared logical volumes.
The copies should reside on separate disks that are controlled by different
disk adapters and are located in separate drawers or units, if possible.
2. Verify the number of logical volume copies by entering: lsvg -l
volume_group_name. In the resulting display, locate the line for the logical
volume for which you just added copies. Notice that the number in the
physical partitions column is x times the number in the logical partitions
column, where x is the number of copies.
3. To verify the placement of logical volume copies, enter: lspv -l hdiskx,
where hdiskx is the name of each disk to which you assigned copies.
Cluster Hardware and Software Preparation 85
Download from Www.Somanuals.com. All Manuals Search And Download.
That is, you enter this command for each disk. In the resulting display,
locate the line for the logical volume for which you just added copies. For
copies placed on separate disks, the numbers in the logical partitions
column and the physical partitions column should be equal. Otherwise, the
copies were placed on the same disk and the mirrored copies will not
protect against disk failure.
Testing a File System
To run a consistency check on each file system’s information:
1. Enter: fsck /filesystem_name
2. Verify that you can mount the file system by entering:
mount /filesystem_name
3. Verify that you can unmount the file system by entering:
umount /filesystem_name
3.4.3 Mirroring Strategies
Shared logical volumes residing on non-RAID disk devices should be
mirrored in AIX to eliminate the disk as a single point of failure. Shared
volume groups residing on a RAID device should not be AIX mirrored; the
disk array provides its own data redundancy.
The copies should reside on separate disks that are controlled by different
disk adapters and are located in separate drawers or units, if possible.
3.4.4 Importing to Other Nodes
The following sections cover: varying off a volume group on the source node,
importing it onto the destination node, changing its startup status, and varying
it off on the destination nodes.
3.4.4.1 Varying Off a Volume Group on the Source Node
After completing the previous tasks, use the varyoffvgcommand to
deactivate the shared volume group. You vary off the volume group so that it
can be properly imported onto a destination node and activated as
appropriate by the cluster event scripts. Enter the following command:
varyoffvg volume_group_name. Make sure that all the file systems of the
volume group have been unmounted, otherwise the varyoffvg command will
not work.
3.4.4.2 Importing a Volume Group onto the Destination Node
This section covers how to import a volume group onto destination nodes
using the SMIT interface. You can also use the TaskGuide utility for this task.
86
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The TaskGuide uses a graphical interface to guide you through the steps of
adding nodes to an existing volume group. For more information on the
Importing the volume group onto the destination nodes synchronizes the
ODM definition of the volume group on each node on which it is imported.
You can use the smit importvgfastpath to import the volume group.
Table 17. smit importvg Options
Options
Description
VOLUME GROUP name
Enter the name of the volume group that you are
importing. Make sure the volume group name is the
same name that you used on the source node.
PHYSICAL VOLUME name
Enter the name of a physical volume that resides in
the volume group. Note that a disk may have a
different logical name on different nodes. Make sure
that you use the disk name as it is defined on the
destination node.
ACTIVATE volume group after it Set the field to yes.
is imported?
Volume
NUMBER
Group
MAJOR If you are not using NFS, you may use the default
(which is the next available number in the valid
range). If you are using NFS, you must make sure to
use the same major number on all nodes. Use the
lvlstmajorcommand on each node to determine a
free major number common to all nodes.
3.4.4.3 Changing a Volume Group’s Startup Status
By default, a volume group that has just been imported is configured to
automatically become active at system restart. In an HACMP for AIX
environment, a volume group should be varied on as appropriate by the
cluster event scripts. Therefore, after importing a volume group, use the SMIT
Change a Volume Group screen to reconfigure the volume group so that it is
not activated automatically at system restart.
Use the smit chvg fastpath to change the characteristics of a volume group.
Table 18. smit crjfs Options
Options
Description
Activate
volume
group Set this field to no.
automatically at system restart?
Cluster Hardware and Software Preparation 87
Download from Www.Somanuals.com. All Manuals Search And Download.
Options
Description
keep the volume group online?
page 88 for a discussion of quorum in an HACMP
cluster.
3.4.4.4 Varying Off the Volume Group on the Destination Nodes
Use the varyoffvg command to deactivate the shared volume group so that it
can be imported onto another destination node or activated as appropriate by
the cluster event scripts. Enter: varyoffvg volume_group_name.
3.4.5 Quorum
Note
This section does not apply to the IBM 7135-110 or 7135-210 RAIDiant
Disk Array, which provides its own data redundancy.
Quorum is a feature of the AIX LVM that determines whether or not a volume
group can be placed online using the varyonvgcommand, and whether or not
it can remain online after a failure of one or more of the physical volumes in
the volume group.
Each physical volume in a volume group has a Volume Group Descriptor
Area (VGDA) and a Volume Group Status Area (VGSA).
VGDA
VGSA
Describes the physical volumes (PVs) and logical volumes (LVs)
that make up a volume group and maps logical partitions to
physical partitions. The varyonvgcommand reads information from
this area.
Maintains the status of all physical volumes and physical partitions
in the volume group. It stores information regarding whether a
physical partition is potentially inconsistent (stale) with mirror
copies on other physical partitions, or is consistent or
synchronized with its mirror copies. Proper functioning of LVM
mirroring relies upon the availability and accuracy of the VGSA
data.
3.4.5.1 Quorum at Vary On
When a volume group is brought online using the varyonvg command, VGDA
and VGSA data structures are examined. If more than half of the copies are
readable and identical in content, quorum is achieved and the varyonvg
88
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
command succeeds. If exactly half the copies are available, as with two of
four, quorum is not achieved and the varyonvgcommand fails.
3.4.5.2 Quorum after Vary On
If a write to a physical volume fails, the VGSAs on the other physical volumes
within the volume group are updated to indicate that one physical volume has
failed. As long as more than half of all VGDAs and VGSAs can be written,
quorum is maintained and the volume group remains varied on. If exactly half
or less than half of the VGDAs and VGSAs are inaccessible, quorum is lost,
the volume group is varied off, and its data becomes unavailable.
Keep in mind that a volume group can be varied on or remain varied on with
one or more of the physical volumes unavailable. However, data contained on
the missing physical volume will not be accessible unless the data is
replicated using LVM mirroring, and a mirror copy of the data is still available
on another physical volume. Maintaining quorum without mirroring does not
guarantee that all data contained in a volume group is available.
Quorum has nothing to do with the availability of mirrored data. It is possible
to have failures that result in loss of all copies of a logical volume, yet the
volume group remains varied on because a quorum of VGDAs/VGSAs are
still accessible.
3.4.5.3 Disabling and Enabling Quorum
Quorum checking is enabled by default. Quorum checking can be disabled
using the chvg -Qn vgnamecommand, or by using the smit chvg fastpath.
Quorum Enabled
With quorum enabled, more than half of the physical volumes must be
available and the VGDA and VGSA data structures must be identical before a
volume group can be varied on with the varyonvgcommand.
With quorum enabled, a volume group will be forced offline if one or more
disk failures cause a majority of the physical volumes to be unavailable.
Having three or more disks in a volume group avoids a loss of quorum in the
event of a single disk failure.
Quorum Disabled
With quorum disabled, all the physical volumes in the volume group must be
available and the VGDA data structures must be identical for the varyonvg
command to succeed. With quorum disabled, a volume group will remain
varied on until the last physical volume in the volume group becomes
unavailable. This section summarizes the effect quorum has on the
availability of a volume group.
Cluster Hardware and Software Preparation 89
Download from Www.Somanuals.com. All Manuals Search And Download.
Forcing a Varyon
A volume group with quorum disabled and one or more physical volumes
unavailable can be “forced” to vary on by using the -f flag with the varyonvg
command. Forcing a varyon with missing disk resources can cause
unpredictable results, including a reducevg of the physical volume from the
volume group. Forcing a varyon should be an overt (manual) action and
should only be performed with a complete understanding of the risks
involved.
The HACMP for AIX software assumes that a volume group is not degraded
and all physical volumes are available when the varyonvg command is issued
at startup or when a volume group resource is taken over during a fallover.
The cluster event scripts provided with the HACMP for AIX software do not
“force” varyon with the -f flag, which could cause unpredictable results. For
this reason, modifying the cluster event scripts to use the -f flag is strongly
discouraged.
Quorum in Non-Concurrent Access Configurations
While specific scenarios can be constructed where quorum protection does
provide some level of protection against data corruption and loss of
availability, quorum provides very little actual protection in non-concurrent
access configurations. In fact, enabling quorum may mask failures by
allowing a volume group to varyon with missing resources. Also, designing
logical volume configuration for no single point of failure with quorum enabled
may require the purchase of additional hardware. Although these facts are
true, you must keep in mind that disabling quorum can result in subsequent
loss of disks—after varying on the volume group—that go undetected.
Quorum in Concurrent Access Configurations
Quorum must be enabled for an HACMP for AIX concurrent access
configuration. Disabling quorum could result in data corruption. Any
concurrent access configuration where multiple failures could result in no
common shared disk between cluster nodes has the potential for data
corruption or inconsistency.
3.4.6 Alternate Method - TaskGuide
The TaskGuide is a graphical interface that simplifies the task of creating a
shared volume group within an HACMP cluster configuration. The TaskGuide
presents a series of panels that guide the user through the steps of specifying
initial and sharing nodes, disks, concurrent or non-concurrent access, volume
group name and physical partition size, and cluster settings. The TaskGuide
can reduce errors, as it does not allow a user to proceed with steps that
90
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
conflict with the cluster’s configuration. Online help panels give additional
information to aid in each step.
3.4.6.1 TaskGuide Requirements
Before starting the TaskGuide, make sure:
• You have a configured HACMP cluster in place.
• You are on a graphics capable terminal.
3.4.6.2 Starting the TaskGuide
You can start the TaskGuide from the command line by typing:
/usr/sbin/cluster/tguides/bin/cl_ccvg or you can use the SMIT interface as
follows:
1. Type smit hacmp.
2. From the SMIT main menu, choose Cluster System Management ->
Cluster Logical Volume Manager ->Taskguide for Creating a Shared
Volume Group. After a pause, the TaskGuide Welcome panel appears.
3. Proceed through the panels to create or share a volume group.
Cluster Hardware and Software Preparation 91
Download from Www.Somanuals.com. All Manuals Search And Download.
92
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 4. HACMP Installation and Cluster Definition
This chapter describes issues concerning the actual installation of HACMP
Version 4.3 and the definition of a cluster and its resources. It concentrates
on the HACMP part of the installation, so, we will assume AIX is already at
the 4.3.2 level. Please refer to the AIX Version 4.3: Migration Guide,
SG24-5116, for details on installation or migration to that level.
This chapter is meant to give an overview of the steps to be taken, and not to
be a complete handbook for performing these tasks. When actually
performing the HACMP install, the HACMP for AIX, Version 4.3: Installation
Guide, SC23-4278 should be consulted.
4.1 Installing HACMP
Before installing, you need to ensure that all the prerequisites are met.
Chapter 8 of the HACMP for AIX, Version 4.3: Installation Guide, SC23-4278,
gives a detailed list. The AIX Level of the server nodes has to be at AIX 4.3.2,
for example, and the required free space in /usrmust be confirmed. For parts
of the product, like HAView, there are prerequisites for other lpps, nv6000 in
this case, that have to be ensured.
You can install either from the installation media, from an installation server
through Network Installation Management (NIM), or from a hard disk to which
the software has been copied.
You will either be installing the HACMP for AIX software for the first time, or
upgrading from an earlier version. Both of those situations are discussed in
the following sections.
4.1.1 First Time Installs
There are a number of filesets involved in an HACMP Installation. Here is a
short overview of them, and what their purpose is.
• cluster.base
This is the basic component that has to be installed on all server nodes in
the cluster, and it contains the following:
cluster.base.client.lib
cluster.base.client.rte
cluster.base.client.utils
cluster.base.server.diag
cluster.base.server.events
cluster.base.server.rte
HACMP Base Client Libraries
HACMP Base Client Runtime
HACMP Base Client Utilities
HACMP Base Server Diags
HACMP Base Server Events
HACMP Base Server Runtime
© Copyright IBM Corp. 1999
93
Download from Www.Somanuals.com. All Manuals Search And Download.
cluster.base.server.utils
HACMP Base Server Utilities
• cluster.cspoc
This component includes all of the commands and environment for the
C-SPOC utility, the Cluster-Single Point Of Control feature. These routines
are responsible for centralized administration of the cluster. There is no
restriction on the node from which you run the C-SPOC utility commands,
so it should also be installed on all the server nodes. It consists of the
following:
cluster.cspoc.rte
cluster.cspoc.cmds
cluster.cspoc.dsh
HACMP CSPOC Runtime Commands
HACMP CSPOC commands
HACMP CSPOC dsh and perl
• cluster.adt
This component contains demo clients and their include files, for example,
for building a clinfo client on a non-AIX machine. Since these are sample
files and demos, you might want to install this on a dedicated machine
only. This machine can further be used for development of server or client
code:
cluster.adt.client.demos
HACMP Client Demos
cluster.adt.client.samples.demos HACMP Client Demos Samples
cluster.adt.client.samples.clinfo HACMP Client clinfo Samples
cluster.adt.client.samples.clstat HACMP Client clstat Samples
cluster.adt.client.include
HACMP Client includes
cluster.adt.client.samples.libcl HACMP Client libcl Samples
cluster.adt.server.samples.images HACMP Sample Images
cluster.adt.server.demos
HACMP Server Demos
cluster.adt.server.samples.demos HACMP Server Sample Demos
• cluster.man.en_US.data
This component contains the man pages in US English. You may like to
exchange this with your own language:
cluster.man.en_US.cspoc.data
cluster.man.en_US.client.data
cluster.man.en_US.server.data
cluster.man.en_US.haview.data
HACMP CSPOC Man pages
HACMP Client Man pages
HACMP Server Man pages
HACMP HAView Man pages
• cluster.msg.en_US
These filesets contain the messages in US English. In contrast to the man
pages, the en_US version must be installed. You might add your
language’s messages if you want:
cluster.msg.en_US.cspoc
cluster.msg.en_US.client
cluster.man.en_US.haview.data
HACMP CSPOC Messages
HACMP Client Messages
HACMP HAView Messages
94
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• cluster.vsm
The Visual Systems Management Fileset contains Icons and bitmaps for
the graphical Management of HACMP Resources, as well as the xhacmpm
command:
cluster.vsm
HACMP X11 Dependent
• cluster.haview
This fileset contains the files for including HACMP cluster views into a
TME 10 Netview Environment. It is installed on a Netview network
management machine, and not on a cluster node:
cluster.haview
HACMP HAView
• cluster.man.en_US.haview.data
This fileset contains man pages and data for the HAView component:
cluster.man.en_US.haview.data
HACMP HAView Manpages
• cluster.msg.en_US.haview
This fileset contains the US English messages for the HAView component:
cluster.msg.en_US.haview
HACMP HAView Messages
Note
TME 10 NetView for AIX must be installed on any system where you will
install HAView. If NetView is installed using a client/server configuration,
HAView should be installed on the NetView client; otherwise, install it on
the NetView server node. Also, be aware that the NetView client should not
be configured as a cluster node to avoid NetView's failure after a failover.
• cluster.taskguides
This is the fileset that contains the taskguide for easy creation of shared
volume groups:
cluster.taskguides.shrvolgrp
HAES Shr Vol Grp Task Guides
• cluster.clvm
This fileset contains the Concurrent Resource Manager (CRM) option:
cluster.clvm
HACMP for AIX Concurrent Access
• cluster.hc
HACMP Installation and Cluster Definition 95
Download from Www.Somanuals.com. All Manuals Search And Download.
This fileset contains the Application Heart Beat Daemon, Oracle Parallel
Server is an application that makes use of it:
cluster.hc.rte
Application Heart Beat Daemon
The installation of CRM requires the following software:
bos.rte.lvm.usr.4.3.2.0
AIX Run-time Executable
Install Server Nodes
From whatever medium you are going to use, install the needed filesets on
each node. Refer to Chapter 8 of the HACMP for AIX, Version 4.3: Installation
Guide, SC23-4278 for details.
Rebooting Servers
The final step in installing the HACMP for AIX software is to reboot each
server in your HACMP for AIX environment.
4.1.2 Upgrading From a Previous Version
If you are upgrading your cluster nodes from a previous version, there are
some things you have to take care of in order to get your existing cluster back
the way you want it after the upgrade is through.
• Ensure that all the prerequisites are met. For details, look into Chapter 8
of the HACMP for AIX Version 4.3: Installation Guide, SC23-4278.
• Archive any localized script and configuration files to prevent losing them
during an upgrade.
• Its always a good idea to have a mksysb of a working system, so take one
of the cluster nodes to be upgraded.
• Commit your current HACMP Version 4.* software (if it is applied but not
committed) so that the HACMP 4.3 software can be installed over the
existing version.
• Save the current configuration using the cluster snapshot utility, and save
any customized event scripts in a directory of your own.
(To review cluster snapshot instructions, see the chapter on saving and
restoring cluster configurations in the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279.)
As the version from where you start may differ, there is more than one way to
get to the required level.
If your site is currently running an earlier version of the HACMP for AIX
software in its cluster environment, except for Version 4.2.2 already running
on AIX 4.3, the following procedures describe how to upgrade your existing
96
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
HACMP software to HACMP for AIX, Version 4.3. The comments on
upgrading the Operating System are not included.
If you are already running AIX 4.3, see the special note at the end of this
section.
Note
Although your objective in performing a migration installation is to keep the
cluster operational and to preserve essential configuration information, do
not run your cluster with mixed versions of the HACMP for AIX software for
an extended period of time.
4.1.2.1 Upgrading from Version 4.1.0 through 4.2.2 to Version 4.3
The following procedure applies to upgrading a two-node or multi-node
cluster running HACMP Version 4.1.0 through 4.2.2 to Version 4.3 when the
installed AIX version is earlier than 4.3.
To perform a rolling AIX migration installation and HACMP upgrade from
Version 4.1.0 through Version 4.2.2 to Version 4.3, complete the following
steps:
Upgrade AIX on One Node
The following steps describe how to upgrade AIX on one node:
1. If you wish to save your cluster configuration, see the chapter “Saving and
Restoring Cluster Configurations” in the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279.
2. Shut down the first node (gracefully with takeover) using the smit clstop
fastpath. For this example, shut down Node A. Node B will take over Node
A's resources and make them available to clients.
See the chapter “Starting and Stopping Cluster Services” in the HACMP
for AIX, Version 4.3: Administration Guide, SC23-4279 for more
information about stopping cluster services.
3. Perform a Migration Installation as described in your AIX Installation
Guide, SBOF-1803 on Node A.
The Migration Installation option preserves the current version of the
HACMP for AIX software and upgrades the existing base operating
system to AIX 4.3.2. Product (application) files and configuration data are
also saved.
4. Check the Migration Installation. Verify that all the disks are available. Run
lppchk -vand oslevel to ensure that the system is in a stable state.
HACMP Installation and Cluster Definition 97
Download from Www.Somanuals.com. All Manuals Search And Download.
Install HACMP 4.3 for AIX on Node A
5. After upgrading AIX and verifying that the disks are correctly configured,
install the HACMP 4.3 for AIX software on Node A. For a short description
Chapter 8 of the HACMP for AIX, Version 4.3: Installation Guide,
SC23-4278.
6. The installation process automatically runs the cl_convertprogram. It
removes the current HACMP objects from /etc/objreposand saves them
to HACMP.old. It creates new HACMP ODM object classes for Version 4.3
in /etc/objrepos.
Note
If upgrading from HACMP 4.1 for AIX, you must run the cl_convertutility
manually. Enter the following command:
/usr/sbin/cluster/conversion/cl_convert -v 4.1
7. Start the Version 4.3 software on Node A using the smit clstartfastpath.
After HACMP is running, start the previous version HACMP software on
Node B, if it is not still running. Check to ensure that the nodes
successfully join the cluster.
Important
If the node running Version 4.3 fails while the cluster is in this state, the
surviving node running the previous version may not successfully mount
the file systems that were not properly unmounted due to Node A’s
failure.
8. Repeat Steps 2 through 7 on Node B on remaining cluster nodes, one at a
time.
Important
In a multi-node cluster, do not synchronize the node configuration or the
cluster topology until the last node has been upgraded.
9. When the last node has been upgraded to both AIX 4.3.2 and HACMP 4.3,
the cluster install/upgrade process is complete.
Check Upgraded Configuration
10.If using tty devices, check that the tty device is configured as a serial
network using the smit chgttyfastpath.
11.In order to verify and synchronize the configuration (if desired), you must
have /.rhosts files on cluster nodes. If they do not exist, create the /.rhosts
98
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
file on Node A using the following command:
/usr/sbin/cluster/utilities/cllsif -x >> /.rhosts
This command will append information to the /.rhosts file instead of
overwriting it. Then, you can ftp this file to the other nodes as necessary.
12.Verify the cluster topology on all nodes using the clverifyutility.
13.Check that custom event scripts are properly installed.
14.Synchronize the node configuration and the cluster topology from Node A
to all nodes (this step is optional).
15.It is recommended that you test the upgraded cluster to ensure proper
behavior.
Client-only Migration
If you are migrating from an HACMP 4.1 for AIX through HACMP 4.2 for AIX
server node to a client-only node running Version 4.3, first remove the
existing server portion of HACMP. If, after upgrading AIX, you install the
cluster.base.client.* filesets on a node running an earlier version of HACMP
for AIX without de-installing the server, the results are unpredictable.
To determine if there is a mismatch between the HACMP client and server
software installed on a node, issue the following command to list the installed
software:
lslpp -L "cluster*"
Examine the list and make sure that all cluster filesets are at 4.3.0.
If you determine that there is a mismatch between the client and server,
de-install the server and then repeat the installation of the client software.
In case the node was just a client before, you only have to install the
cluster.base.client.* filesets after it has been migrated to AIX Version 4.3.2.
Again, please check whether the installation succeeded by issuing the
command: lslpp -L "cluster*"
4.1.2.2 Upgrading from Version 4.2.2 on AIX 4.3.2 to HACMP Version 4.3
The following steps describe upgrading from Version 4.2.2 on AIX 4.3.2 to
HACMP Version 4.3:
1. If your cluster is currently running Version 4.2.2 of the HACMP software on
AIX Version 4.3.2, you should upgrade your cluster configuration to
HACMP 4.3 for AIX.
HACMP Installation and Cluster Definition 99
Download from Www.Somanuals.com. All Manuals Search And Download.
2. If you wish to save your cluster configuration, see the chapter Saving and
Restoring Cluster Configurations in the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279.
3. Commit your current HACMP for AIX software on all nodes.
4. Shut down one node (gracefully with takeover) using the smit clstop
fastpath. For this example, shut down Node A. Node B will take over Node
A’s resources and make them available to clients.
See the chapter “Starting and Stopping Cluster Services” in the HACMP
for AIX, Version 4.3: Administration Guide, SC23-4279, for more
information on stopping cluster services.
5. Install HACMP for AIX Version 4.3. See Chapter 8 of the HACMP for AIX,
Version 4.3: Installation Guide, SC23-4278, starting with the section
“Installation Choices”, for instructions.
The cl_convertutility automatically updates the HACMP ODM object
classes to the 4.3 version.
Note
If IP address swapping is being used on this node, that is, a boot
address is defined for this node, check to ensure that the HACMP
changes to /etc/inittaband /etc/rc.netexist as specified in Appendix
A of the HACMP for AIX, Version 4.3: Installation Guide, SC23-4278,
before rebooting the node.
6. Reboot Node A.
7. Start the HACMP for AIX software on Node A using the smit clstart
fastpath and verify that Node A successfully joins the cluster.
8. Repeat Steps 3 through 7 on remaining cluster nodes, one at a time.
9. After all nodes have been upgraded to HACMP Version 4.3, synchronize
the node configuration and the cluster topology from Node A to all nodes.
10.Verify the cluster topology on all nodes using the clverifyutility.
11.Complete a test phase on the cluster before putting it into production.
4.2 Defining Cluster Topology
The cluster topology is comprised of the following components:
• The cluster definition
• The cluster nodes
• The network adapters
100
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• The network modules
You define the cluster topology by entering information about each
component into HACMP-specific ODM classes. You enter the HACMP ODM
data by using the HACMP SMIT interface or the VSM utility xhacmpm. The
xhacmpmutility is an X Windows tool for creating cluster configurations using
icons to represent cluster components. For more information about the
xhacmpmutility, see the administrative facilities chapter of the HACMP for AIX,
Version 4.3: Concepts and Facilities, SC23-4276.
Note
The SP Switch network module can support multiple clusters; therefore, its
settings should remain at their default values to avoid affecting HACMP
event scripts. If you must change these settings, see the chapter on
changing the cluster topology in the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279 for more information.
4.2.1 Defining the Cluster
The cluster ID and name identifies a cluster in an HACMP environment. The
cluster ID and name must be unique for each cluster defined.
Cluster IDs have to be a positive integer in the range from 1 through 99999,
and the cluster name is a text string of up to 31 alphanumeric characters,
including underscores. It doesn’t necessarily need to match the hostname.
The HACMP software uses this information to create the cluster entries for
the ODM.
4.2.2 Defining Nodes
After defining the cluster name and ID, cluster nodes have to be defined. As
above, this is usually done through smit hacmp. Each of the cluster nodes
needs a unique name, so the cluster manager can address them.
Again, a node name is a text string of up to 31 alphanumeric characters that
can contain underscores.
You can add more than one node at a time by separating them with
whitespace characters.
HACMP Installation and Cluster Definition 101
Download from Www.Somanuals.com. All Manuals Search And Download.
Note
The node names are logically sorted in their ascii order within HACMP in
order to decide which nodes are considered to be neighbors for heartbeat
purposes.
In order to build a logical ring, a node always talks to its up- and
downstream neighbor in their node name’s ascii order. The uppermost and
the lowest node are also considered neighbors.
Adding or Changing a Node Name after the Initial Configuration
If you want to add or change a node name after the initial configuration, use
the Change/Show Cluster Node Name screen. See the chapter on changing
the cluster topology of the HACMP for AIX, Version 4.3:Administration Guide,
SC23-4279 for more information.
4.2.3 Defining Adapters
To define the adapters after defining the node names, first consult your
planning worksheets for both TCP/IP and serial networks.
There are a number of attributes associated with an Adapter in the HACMP
configuration which need to be specified:
Adapter IP Label
Enter the IP label (the name) of the adapter you have
chosen as the service address for this adapter. Adapter
labels can be any ASCII text string consisting of
alphabetical and numeric characters, underscores, and
hyphens, up to 31 characters.
If IP address takeover is defined for that adapter, a boot
adapter (address) label has to be defined for it. Use a
consistent naming convention for boot adapter labels.
(You will choose the Add an Adapter option again to
define the boot adapter when you finish defining the
service adapter.)
You can use hyphens in adapter labels. However,
currently it might not be a good idea since the
/usr/sbin/cluster/diag/clverifyutility flags adapter
labels that contain hyphens each time it runs.
Network Type
Indicate the type of network to which this adapter is
connected. Pre-installed network modules are listed on
the pop-up pick list.
102
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Network Name
Enter an ASCII text string that identifies the network.
The network name can include alphabetic and numeric
characters and underscores. Use no more than 31
characters. The network name is arbitrary, but must be
used consistently for adapters on the same physical
network.
If several adapters share the same physical network,
make sure you use the same network name for each of
these adapters.
Network Attribute Indicate whether the network is public, private, or
serial. Press Tab to toggle the values. In the context of
HACMP, serial networks means “non-TCP/IP”; public
and private networks are TCP/IP networks. Ethernet,
Token-Ring, FDDI, and SLIP are public networks.
SOCC, ATM, and an SP Switch are private networks.
RS232 lines, target mode SSA loops, and target mode
SCSI-2 buses are serial networks.
Adapter Function Indicate whether the adapter's function is service,
standby, or boot. Press Tab to toggle the values. A node
has a single service adapter for each public or private
network. A serial network has only a single service
adapter.
A node can have none, one, or more standby adapters
for each public network. Serial and private networks do
not have standby adapters, with the exception of ATM
networks. ATM networks must be defined as private,
and therefore standby adapters are supported.
In an HACMP environment on the RS/6000 SP, the
ethernet adapters can be configured as service
adapters but should not be configured for IP address
takeover. Regarding the SP Switch, network, boot, and
service addresses used for IP address takeover are
ifconfig alias addresses used on the css0 network. See
the appendix entitled HACMP for the RS/6000 SP in the
HACMP Installation Guide Version 4.3 for more
information on adapter functions in an SP Switch
environment.
In an ATM network, the adapter function should be listed
as svc_s to indicate that the interface is used by
HACMP servers. Keep in mind that the netmask for all
adapters in an HACMP network must be the same.
HACMP Installation and Cluster Definition 103
Download from Www.Somanuals.com. All Manuals Search And Download.
Adapter Identifier Enter the IP address in dotted decimal format or a
device file name. IP address information is required for
non-serial network adapters only if the node’s address
cannot be obtained from the domain name server or the
local /etc/hosts file (using the adapter IP label given).
You must enter device filenames for serial network
adapters. RS232 serial adapters must have the device
filename /dev/ttyN. Target mode SCSI serial adapters
must have the device file name /dev/tmscsiN. Target
mode SSA adapters must have the device file name
/dev/tmssaN.im or /dev/tmssaN.tm.
Adapter Hardware Address(optional) Enter a hardware address for the
adapter. The hardware address must be unique within
the physical network. Enter a value in this field only if:
You are currently defining a service adapter, and the
adapter has a boot address, and you want to use
hardware address swapping. See the chapter on
planning TCP/IP networks in the HACMP for AIX,
Version 4.3:Planning Guide, SC23-4277 for more
information on hardware address swapping. This facility
is supported only for Ethernet, Token Ring, and FDDI
adapters. It does not work with the SP Switch.
Node Name
Define a node name for all adapters except for those
service adapters whose addresses may be shared by
nodes participating in the resource chain for a rotating
resource configuration. These adapters are rotating
resources. The event scripts use the user-defined
configuration to associate these service addresses with
the proper node. In all other cases, addresses are
associated with a particular node (service, boot, and
standby)
Note
Although it is possible to have only one physical network adapter (no
standby adapters), this constitutes a potential single point of failure
condition and is not recommended for an HACMP configuration.
104
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Note
When IPAT is configured, the run level of the IP-related entries (e. g.
rctcpip, rcnfs...) of the /etc/inittabare changed to “a”. This has the result
that these services are not started at boot time, but with HACMP.
Adding or Changing Adapters after the Initial Configuration
If you want to change the information about an adapter after the initial
configuration, use the Change/Show an Adapter screen. See the chapter on
changing the cluster topology in the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279, for more information.
4.2.4 Configuring Network Modules
Each supported cluster network in a configured HACMP cluster has a
corresponding cluster network module. Each network module monitors all I/O
to its cluster network.
The Network Modules are pre-loaded when you install the HACMP software.
You do not need to enter information in the Network Module SMIT screens
unless you want to change some field associated with a network module,
such as the failure detection rate.
Each network module maintains a connection to other network modules in the
cluster. The Cluster Managers on cluster nodes send messages to each other
through these connections. Each network module is responsible for
maintaining a working set of service adapters and for verifying connectivity to
cluster peers. The network module is also responsible for reporting when a
given link actually fails. It does this by sending and receiving periodic
heartbeat messages to or from other network modules in the cluster, and
reporting back to the Cluster Manager when it misses a threshold number of
heartbeats.
Currently, network modules support communication over the following types
of networks:
• Serial (RS232)
• Target-mode SCSI
• Target-mode SSA
• IP (Generic IP)
• Ethernet
• Token-Ring
• FDDI
• SOCC
HACMP Installation and Cluster Definition 105
Download from Www.Somanuals.com. All Manuals Search And Download.
• SLIP
• SP Switch
• ATM
It is highly unlikely that you will add or remove a network module. For
information about changing a characteristic of a Network Module, such as the
failure detection rate, see the chapter on changing the cluster topology in the
HACMP for AIX, Version 4.3: Administration Guide, SC23-4279. Changing
the network module allows the user to influence the rate of heartbeats being
sent and received by a Network Module, thereby changing the sensitivity of
the detection of a network failure.
In HACMP/ES, topology services and group services are used instead of
Network Interface Modules (NIMs) in order to keep track of the status of
nodes, adapters or resources.
In HACMP/ES, the tuning of network sensitivity is a little different.
Customizable attributes are the interval between heartbeats in seconds and
the Fibrillate Count, which is the acceptable number of missed heartbeats
before some event is triggered. You will find the Change / Show Topology
and Group Services Configuration in the Cluster Topology screen, just
like the NIM tuning options.
4.2.5 Synchronizing the Cluster Definition Across Nodes
Synchronization of the cluster topology ensures that the ODM data on all
cluster nodes is in sync. The HACMP ODM entries must be the same on each
node in the cluster. If the definitions are not synchronized across nodes, the
HACMP for AIX software generates a run-time error at cluster startup.
Even if you have a cluster defined with only one node, you must still
synchronize the cluster.
The processing performed in synchronization varies depending on whether
the cluster manager is active on the local node. If the cluster manager is not
active on the local node when you select this option, the ODM data in the
system default configuration directory (DCD) on the local node is copied to
the ODMs stored in the DCDs on all cluster nodes. The cluster manager is
typically not running when you synchronize the initial cluster configuration.
If the cluster manager is active on the local node, the ODM data stored in the
DCDs on all cluster nodes are synchronized. In addition, the configuration
data stored in the active configuration directory (ACD) on each cluster node is
overwritten with the new configuration data, which becomes the new active
106
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
configuration. If the cluster manager is active on some other cluster nodes
but not on the local node, the synchronization operation is aborted.
Before attempting to synchronize a cluster configuration, ensure that all
nodes are powered on, that the HACMP software is installed, and that the
/etc/hosts and /.rhosts files on all nodes include all HACMP boot and service
IP labels.
The /.rhosts file may not be required if you are running HACMP on the SP
system.The SP system uses kerberos as its security infrastructure. If you are
running HACMP on a node with kerberos enabled (usually an SP node, but
could also be a standalone RS/6000 that has been configured with kerberos),
you can set a parameter in HACMP to use “Enhanced Security”. This feature
removes the requirement of TCP/IP access control lists (for example, the
/.rhosts file) on remote nodes during HACMP configuration. Instead, it uses a
kerberized version of remote commands to accomplish the synchronization.
It should be noted that Kerberos support is not included in standard AIX
4.3.2. However, Kerberos is available in the public domain, and it is possible
to get it and configure it on a non-SP RS/6000 node. This is not very common
though, so you will almost always see HACMP Enhanced Security used on
the SP system.
When you synchronize the cluster topology, there are two options that control
the behavior of this process as follows:
Table 19. Options for Synchronization of the Cluster Topology
Ignore Cluster
Verification Errors
If you specify yes, the result of the cluster verification
performed as part of synchronization is ignored and the
configuration is synchronized even if verification fails.
If you specify no, the changes are not synchronized if
verification fails. View the error messages in the system
error log to determine the configuration problem.
For information about the
/usr/sbin/cluster/diag/clverifyutility, see the
chapter on verifying a cluster configuration in the HACMP
for AIX, Version 4.3: Administration Guide, SC23-4279.
Emulate or Actual
If you set this field to Emulate, the synchronization will be
an emulation and will not affect the Cluster Manager. If you
set this field to Actual, the synchronization will actually
occur, and any subsequent changes will be made to the
Cluster Manager. Emulate is the default value.
The cluster topology definition (including all node, adapter, and network
module information) is copied to the other nodes in the cluster.
HACMP Installation and Cluster Definition 107
Download from Www.Somanuals.com. All Manuals Search And Download.
4.3 Defining Resources
The HACMP for AIX software provides a highly available environment by
identifying a set of cluster-wide resources essential to uninterrupted
processing, and then by defining relationships among nodes that ensure
these resources are available to client processes. Resources include the
following hardware and software:
• Disks
• Volume groups
• File systems
• Network addresses
• Application servers
In the HACMP for AIX software, you define each resource as part of a
resource group. This allows you to combine related resources into a single
logical entity for easier configuration and management. You then configure
each resource group to have a particular kind of relationship with a set of
nodes. Depending on this relationship, resources can be defined as one of
Group Options” on page 28 for details.
After configuring the cluster topology, you must configure resources and set
up the cluster node. This involves:
• Configuring resource groups and node relationships to behave as desired
• Adding individual resources to each resource group
• Setting up run-time parameters for each node
• Synchronizing cluster nodes
4.3.1 Configuring Resource Groups
Resource Groups are initialized by telling the HACMP ODM their names, the
participating nodes, and their relationship. The order in which the
participating nodes are defined is taken as the priority of the resource chain,
that is, priority is decreasing from left to right.
The relationship can be one of Cascading, Rotating or Concurrent. See 2.4.1,
“Resource Group Options” on page 28 for details.
108
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
4.3.1.1 Configuring Resources for Resource Groups
Once you have defined resource groups, you further configure them by
assigning cluster resources to one resource group or another. You can
configure resource groups even if a node is powered down. However, SMIT
cannot list possible shared resources for the node (making configuration
errors likely).
Note
You cannot configure a resource group until you have completed the
information on the Add a Resource Group screen.
Note
If you configure a cascading resource group with an NFS mount point, you
must also configure the resource to use IP Address Takeover. If you do not
do this, takeover results are unpredictable.You should also set the field
value Filesystems Mounted Before IP Configured to true so that the
takeover process proceeds correctly.
Note
When setting up a cascading resource with an IP Address takeover
configuration, each cluster node should be configured in no more than
(N+1) resource groups on a particular network. Here, N is the number of
standby adapters on a particular node and network.
The following describes the different possibilities for resources that might be
added to a resource group:
Table 20. Options Configuring Resources for a Resource Group
Service IP Label
If IP address takeover is being used, list the IP label to be
moved when this resource group is taken over. Press F4
to see a list of valid IP labels. These include addresses
which rotate or may be taken over.
HTY Service IP Label
File Systems
NTX adapters are not supported by HACMP for AIX 4.3
Identify the file systems to include in this resource group.
Press F4 to see a list of the file systems. When you enter
a file system in this field, the HACMP for AIX software
determines the correct values for the Volume Groups and
Raw Disk PVIDs fields.
HACMP Installation and Cluster Definition 109
Download from Www.Somanuals.com. All Manuals Search And Download.
Service IP Label
If IP address takeover is being used, list the IP label to be
moved when this resource group is taken over. Press F4
to see a list of valid IP labels. These include addresses
which rotate or may be taken over.
File Systems
Identify the method for checking consistency of file
Consistency Check
systems, fsck(default) or logredo(for fast recovery).
File Systems Recovery Identify the recovery method for the file systems, parallel
Method
(for fast recovery) or sequential (default).
Do not set this field to parallel if you have shared, nested
file systems. These must be recovered sequentially. (Note
that the cluster verification utility, clverify, does not report
file system and fast recovery inconsistencies.)
File Systems to Export
Identify the file systems to be exported to include in this
resource group. These should be a subset of the file
systems listed above. Press F4 for a list.
File Systems to
NFS Mount
Identify the subset of file systems to NFS mount. All nodes
in the resource chain that do not currently hold the
resource will attempt to NFS mount these file systems
while the owner node is active in the cluster.
These settings also have to be synchronized throughout the cluster.
Therefore Synchronize Cluster Resources has to be chosen from the
corresponding SMIT Menu.
If the Cluster Manager is running on the local node, synchronizing cluster
resources triggers a dynamic reconfiguration event (DARE, see 8.5.3, “DARE
4.3.1.2 Configuring Run-Time Parameters
There are two types of Run-Time Parameters for a node that can be chosen.
One of them is the debug level, which can be switched from high to low,
meaning all cluster manager actions are logged, or only errors are logged,
respectively. The other is the differentiation whether the node uses NIS or
DNS nameservice or not, to enable the cluster manager to turn that off in
case it would interfere with its actions. Both of these parameters can be
changed while the cluster is running.
4.3.1.3 Defining Application Servers
Application servers are another resource that can be configured into a
Resource Group. They consist of a (hopefully meaningful) name, in order to
enable the cluster manager to identify the application server uniquely, as well
110
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
as the path locations for start and stop scripts for the application. These
scripts have to be in the same location on every service node.
Just as for pre- and post-events, these scripts can be adapted to specific
nodes. They don’t need to be equal in content. The system administrator has
to ensure, however, that they are in the same location, use the same name,
and are executable for the root user.
4.3.1.4 Synchronizing Cluster Resources
After defining these resources and their relationship with the resource
group, the act of synchronizing cluster resources sends the information
contained on the current node to all defined cluster nodes.
Note
All configured nodes must be on their boot addresses when a cluster has
been configured and the nodes are synchronized for the first time. Any
node not on its boot address will not have its /etc/rc.net file updated with
the HACMP entry; this causes problems for the reintegration of this node
into the cluster.
If a node attempts to join the cluster when its configuration is out-of-sync with
other active cluster nodes, it will be denied. You must ensure that other nodes
are synchronized to the joining member.
4.4 Initial Testing
After installing and configuring your cluster, it is recommended that you do
some initial testing in order to verify that the cluster is acting as it should.
4.4.1 Clverify
Running /usr/sbin/cluster/diag/clverifyis probably a good start to the
testing. It allows you to check the software and the cluster.
Software checking is reduced to lpp checking, which is basically checking
whether HACMP-specific modifications to AIX files are correct. For
correctness of the installation itself, use the lppcheck -v command.
Cluster verification is divided into topology and configuration checking. These
two parts do basically the same as smit clverify, i.e. verifying that the
clusters topology as well as the resource configurations are in sync on the
cluster nodes.
HACMP Installation and Cluster Definition 111
Download from Www.Somanuals.com. All Manuals Search And Download.
4.4.2 Initial Startup
At this point in time, the cluster is not yet started. So the cluster manager has
to be started first. To check whether the cluster manager is up, you can either
look for the process with the pscommand:
ps -ef | grep clstr
or look for the status of the cluster group subsystems:
lssrc -g cluster
or look for the status of the network interfaces. If you have IP Address
Takeover (IPAT) configured you should see that the network interface is on its
boot address with the netstat -i command.
Then start HACMP through smit clstart. In the panel that appears, choose
the following parameters and press Enter:
1. start now
2. broadcast message true
3. start cluster lock services false
4. start cluster information daemon true
Reissue either the pscommand (see above) or look for the interface state
with the netstat -i command. Now, you should see that the boot interface is
gone in favor of the service-interface.
You also would like to check whether a takeover will work, so, you have to
bring up HACMP on all cluster nodes through smitty clstartand check
whether the cluster gets into a stable state. Use clstatfor this purpose.
4.4.3 Takeover and Reintegration
When the cluster is up and running, stop one of the node’s cluster managers
with smitty clstopand choose graceful with takeover. One possibility to
check whether the takeover went through smoothly is to look at the
/tmp/hacmp.out file during the takeover, preferably on the takeover node. You
can use the tail -f /tmp/hacmp.outcommand for this.
After the cluster has become stable, you might check the netstat -ioutput
again to verify that the takeover node has acquired the IP address of the
“failed” node.
112
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
For cascading resource groups the failed node is going to reaquire its
resources, once it is up and running again. So, you have to restart HACMP on
it through smitty clstartand check again for the logfile, as well as the
clusters status.
Further and more intensive debugging issues are covered in Chapter 7,
4.5 Cluster Snapshot
Now that the actual installation is finished, the cluster is well documented in
the planning sheets, all information from there has been implemented in the
HACMP ODM, and the cluster is verified and synchronized; provided the
initial testing didn’t bring up any curiosities, you should save this working
configuration in a cluster snapshot.
The cluster snapshot utility allows you to save, in a file, a record of all the
data that defines a particular cluster configuration. This facility gives you the
ability to recreate a particular cluster configuration, a process called applying
a snapshot, provided the cluster is configured with the requisite hardware and
software to support the configuration.
You can perform many of the cluster snapshot utility operations, such as
saving a configuration and applying a saved configuration, using the HACMP
for AIX VSM application (xhacmpm). For more information, see the
administrative facilities chapter in the HACMP for AIX, Version 4.3: Concepts
and Facilities, SC23-4276, and the online help information available with the
application.
In addition, a snapshot can provide useful information for troubleshooting
cluster problems. Because the snapshots are simple ASCII files that can be
sent via e-mail, they can make remote problem determination easier.
You can also add your own custom snapshot methods to store additional
user-specified cluster and system information in your snapshots. The output
from these user-defined custom methods is reported along with the
conventional snapshot information.
Note
You cannot use the cluster snapshot facility in a cluster with nodes
concurrently running different versions of HACMP for AIX.
HACMP Installation and Cluster Definition
113
Download from Www.Somanuals.com. All Manuals Search And Download.
Essentially, a snapshot saves all the ODM classes HACMP has generated
during its configuration. It does not save user customized scripts, such as
start or stop scripts for an application server. However, the location and
names of these scripts are in an HACMP ODM class, and are therefore
saved. It is very helpful to put all the customized data in one defined place, in
order to make saving these customizations easier. You can then use a
custom snapshot method to save this data as well, by including a
user-defined script in the custom snapshot.
4.5.1 Applying a Cluster Snapshot
Applying a cluster snapshot overwrites the data in the existing HACMP for
AIX ODM classes on all nodes in the cluster with the new ODM data
contained in the snapshot. You can apply a cluster snapshot from any cluster
node. However, you have to differentiate between two possible states the
cluster could be in when applying the snapshot.
If cluster services are inactive on all cluster nodes, applying the snapshot
changes the ODM data stored in the system default configuration directory
(DCD). If cluster services are active on the local node, applying a snapshot
triggers a cluster-wide dynamic reconfiguration event. In dynamic
reconfiguration, in addition to synchronizing the ODM data stored in the
DCDs on each node, HACMP for AIX replaces the current configuration data
stored in the active configuration directory (ACD) with the changed
configuration data in the DCD. The snapshot becomes the currently active
configuration.
Note
A cluster snapshot used for dynamic reconfiguration may contain changes
to either the cluster topology OR to cluster resources, but not both. You
cannot change both the cluster topology and cluster resources in a single
dynamic reconfiguration event.
Note
Applying a cluster snapshot may affect both AIX and HACMP for AIX ODM
objects and system files as well as user-defined files.
More detailed Information about Cluster Snapshot can be found in the
HACMP for AIX, Version 4.3: Administration Guide, SC23-4279, Chapter 11
as well as in the HACMP for AIX, Version 4.3: Troubleshooting Guide,
SC23-4280.
114
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
HACMP Installation and Cluster Definition
115
Download from Www.Somanuals.com. All Manuals Search And Download.
116
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 5. Cluster Customization
Within an HACMP for AIX cluster, there are several things that are
customizable. The following paragraphs explain the customizing features for
events, error notification, network modules and topology services.
5.1 Event Customization
An HACMP for AIX cluster environment acts upon a state change with a set
of predefined cluster events (see 5.1.1, “Predefined Cluster Events” on page
117). Whenever a state change is detected by the cluster manager, it decides
which event will be started. It then executes the script for that event in a shell,
as well as the subevents associated with it. These predefined events can be
found under /usr/sbin/cluster/events.
The HACMP for AIX software provides an event customization facility that
allows you to tailor event processing to your site. This facility can be used to
include the following types of customization:
• Adding, changing, and removing custom cluster events
• Pre- and post-event processing
• Event notification
• Event recovery and retry
5.1.1 Predefined Cluster Events
HACMP has the following predefined cluster events:
5.1.1.1 Node Events
This is the sequence of node_up events:
node_up
This event occurs when a node joins the cluster.
Depending on whether the node is local or
remote, this event initiates either a
node_up_local or node_up_remote event.
node_up_local
This script acquires the service address (or
shared address), gets all its owned (or shared)
resources, and takes the resources. This
includes making disks available, varying on
volume groups, mounting file systems, exporting
file systems, NFS-mounting file systems, and
varying on concurrent access volumes groups.
© Copyright IBM Corp. 1999
117
Download from Www.Somanuals.com. All Manuals Search And Download.
acquire_service_addr
acquire_takeover_addr
(If configured for IP address takeover.)
Configures boot addresses to the corresponding
service address, and starts TCP/IP servers and
network daemons by running the telinit -a
command.
The script checks to see if a configured standby
address exists, then swaps the standby address
with the takeover address.
get_disk_vg_fs
Acquires disk, volume group, and file system
resources.
node_up_remote
Causes the local node to release all resources
taken from the remote node and to place any
concurrent volume groups in concurrent mode.
Some of the scripts called by node_up_remote
include the following:
release_takeover_addr
(If configured for IP address takeover.) Identifies
a takeover address to be released because a
standby adapter on the local node is
masquerading as the service address of the
remote node. Reconfigures the local standby
adapter to its original address (and hardware
address, if necessary).
stop_server
Stops application servers belonging to the
reintegrating node.
release_vg_fs
Releases volume groups and file systems
belonging to a resource group that the remote
node will be taking over.
cl_deactivate_nfs
node_up_complete
Unmounts NFS file systems.
This event occurs only after a node_up event
has successfully completed. Depending on
whether the node is local or remote, this event
initiates either a node_up_local_complete or
node_up_remote_complete event.
node_up_local_complete Calls the start_server script to start application
servers. This event occurs only after a
node_up_local event has successfully
completed.
node_up_remote_completeAllows the local node to do an NFS mount only
after the remote node is completely up. This
118
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
event occurs only after a node_up_remote event
has successfully completed.
Sequence of node_down Events
node_down
This event occurs when a node intentionally
leaves the cluster or fails. Depending on
whether the exiting node is local or remote, this
event initiates either the node_down_local or
node_down_remote event, which in turn initiates
a series of subevents.
node_down_local
stop_server
Processes the following events:
Stops application servers.
release_takeover_addr
(If configured for IP address takeover.) Identifies
a takeover address to be released because a
standby adapter on the local node is
masquerading as the service address of the
remote node. Reconfigures the local standby
with its original IP address (and hardware
address, if necessary).
release_vg_fs
Releases volume groups and file systems that
are part of a resource group the local node is
serving.
release_service_addr
(If configured for IP address takeover.)
Detaches the service address and reconfigures
the service adapter to its boot address.
node_down_remote
Processes the following events:
acquire_takeover_addr
(If configured for IP address takeover.) Checks
for a configured standby address currently seen
as up by the Cluster Manager, and then does a
standby_address to takeover_address swap
(and hardware address, if necessary.
get_disk_vg_fs
Acquires disk, volume group, and file system
resources as part of a takeover.
node_down_complete
This event occurs only after a node_down event
has successfully completed. Depending on
whether the node is local or remote, this event
initiates either a node_down_local_complete or
node_down_remote_complete event.
Cluster Customization
119
Download from Www.Somanuals.com. All Manuals Search And Download.
node_down_local_completeInstructs the Cluster Manager to exit when the
local node has left the cluster. This event occurs
only after a node_down_local event has
successfully completed.
node_down_remote_completeStarts takeover application servers. This
event runs only after a node_down_remote
event has successfully completed.
start_server
Starts application servers.
5.1.1.2 Network Events
network_down
This event occurs when the Cluster Manager
determines a network has failed. A
network_down event can take one of two forms:
Local network_down, where only a particular
node has lost contact with a network.
Global network_down, where all of the nodes
connected to a network have lost contact with a
network. It is assumed in this case that a
network-related failure has occurred rather than
a node-related failure.
The network_down event mails a notification to
the system administrator, but takes no further
action since appropriate actions depend on the
local network configuration.
network_down_complete This event occurs only after a network_down
event has successfully completed. The default
network_down_complete event processing
takes no actions since appropriate actions
depend on the local network configuration.
network_up
This event occurs when the Cluster Manager
determines a network has become available for
use. The default network_up event processing
takes no actions since appropriate actions
depend on the local network configuration.
network_up_complete
This event occurs only after a network_up event
has successfully completed. The default
network_up_complete event processing takes
120
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
no actions since appropriate actions depend on
the local network configuration.
5.1.1.3 Network Adapter Events
swap_adapter
This event occurs when the service adapter on a
node fails. The swap_adapter event exchanges
or swaps the IP addresses of the service and a
standby adapter on the same HACMP network
and then reconstructs the routing table.
swap_adapter_complete This event occurs only after a swap_adapter
event has successfully completed. The
swap_adapter_complete event ensures that the
local ARP cache is updated by deleting entries
and pinging cluster IP addresses.
fail_standby
join_standby
This event occurs if a standby adapter fails or
becomes unavailable as the result of an IP
address takeover. The fail_standby event
displays a console message indicating that a
standby adapter has failed or is no longer
available.
This event occurs if a standby adapter becomes
available. The join_standby event displays a
console message indicating that a standby
adapter has become available.
5.1.1.4 Cluster Status Events
config_too_long
This event occurs when a node has been in
reconfiguration for more than six minutes. The
event periodically displays a console
message.
reconfig_topology_start
This event marks the beginning of a dynamic
reconfiguration of the cluster topology.
reconfig_topology_completeThis event indicates that a cluster topology
dynamic reconfiguration has completed.
reconfig_resource_acquire This event indicates that cluster resources
that are affected by dynamic reconfiguration
are being acquired by appropriate nodes.
reconfig_resource_release This event indicates that cluster resources
affected by dynamic reconfiguration are being
released by appropriate nodes.
Cluster Customization
121
Download from Www.Somanuals.com. All Manuals Search And Download.
reconfig_resource_completeThis event indicates that a cluster resource
dynamic reconfiguration has completed.
5.1.2 Pre- and Post-Event Processing
To tailor event processing to your environment, specify commands or
user-defined scripts that should execute before and/or after a specific event
is generated by the Cluster Manager. You specify them by selecting the
HACMP event to be customized on the smit hacmp -> Cluster
Configuration -> Resources -> Cluster Events -> Change/Show Cluster
Events screen, and then, choosing the one to be tailored. Now, you can
enter the location of your pre- or post-event to be executed before or after the
chosen event has been processed.
For preprocessing, for example, you may want to send a message to specific
users informing them to stand by while a certain event occurs. For
post-processing, you may want to disable login for a specific group of users if
a particular network fails.
5.1.3 Event Notification
You can specify a command or user-defined script that provides notification
(for example, mail) that an event is about to happen and that an event has
just occurred, along with the success or failure of the event.
This is done on the very same SMIT screen as in 5.1.2, “Pre- and Post-Event
For example, a site may want to use a network_downnotification event to
inform system administrators that traffic may have to be rerouted. Afterwards,
you can use a network_upnotification event to tell system administrators that
traffic can again be serviced through the restored network.
Event notification in an HACMP cluster can also be done using pre- and
post-event scripts, just by adding the script you want to execute for
notification into the pre-and/or post-eventcommand script.
5.1.4 Event Recovery and Retry
You can specify a command that attempts to recover from an event command
failure. If the retry count is greater than zero, and the recovery command
succeeds, the event script command is rerun. You can also specify the
number of times to attempt to execute the recovery command.
122
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
For example, a file system cannot be unmounted, because of a process
running on it. Then, you might want to kill that process first, before
unmounting the file system, in order to get the event script done. Now, since
the event script didn’t succeed in its first run, the Retry feature enables
HACMP for AIX to retry it until it finally succeeds, or the retry count is
reached.
5.1.5 Notes on Customizing Event Processing
You must declare a shell (for example #!/bin/sh) at the beginning of each
script executed by the notify, recovery, and pre- or post-event processing
commands.
Notify, recovery, and pre- and post-event processing do not occur when the
force option of the node_downevent is specified.
Synchronizing the cluster configuration does not propagate the actual new or
changed scripts; you must add these to each node manually. Also, it is
allowed to have different contents in these scripts on different nodes in order
to be able to act upon different environments. However, the name of these
scripts, their location in the file system, and their permission bits have to be
identical.
5.1.6 Event Emulator
To test the effect of running an event on your cluster, HACMP for AIX
provides a utility to run an emulation of an event. This emulation lets you
predict a cluster's reaction to an event as though the event actually occurred.
The emulation runs on all active nodes in your cluster, and the output is
stored in an output file. You can select the path and name of this output file
using the EMU_OUTPUTenvironment variable, or, use the default
/tmp/emuhacmp.out file on the node that invoked the Event Emulator.
For more information on event emulation, see these chapters: “Administrative
Facilities” in the HACMP for AIX, Version 4.3: Concepts and Facilities,
SC23-4276, and “Monitoring an HACMP Cluster” in the HACMP for AIX,
Version 4.3: Administration Guide, SC23-4279.
5.2 Error Notification
The AIX Error Notification facility detects errors matching predefined
selection criteria and responds in a programmed way. The facility provides a
wide range of criteria that you can use to define an error condition. These
errors are called notification objects.
Cluster Customization
123
Download from Www.Somanuals.com. All Manuals Search And Download.
Each time an error is logged in the system error log, the error notification
daemon determines if the error log entry matches the selection criteria. If it
does, an executable is run. This executable, called a notify method, can
range from a simple command to a complex program. For example, the notify
method might be a mail message to the system administrator or a command
to shut down the cluster.
Using the Error Notification facility adds an additional layer of high availability
to the HACMP for AIX software. Although the combination of the HACMP for
AIX software and the inherent high availability features built into the AIX
operating system keeps single points of failure to a minimum, failures still
exist that, although detected, are not handled in a useful way.
Take the example of a cluster where an owner node and a takeover node
share an SCSI disk. The owner node is using the disk. If the SCSI adapter on
the owner node fails, an error may be logged, but neither the HACMP for AIX
software nor the AIX Logical Volume Manager responds to the error. If the
error has been defined to the Error Notification facility, however, an
executable that shuts down the node with the failed adapter could be run,
allowing the surviving node to take over the disk.
5.3 Network Modules/Topology Services and Group Services
The HACMP for AIX SMIT interface allows you to add, remove, or change an
HACMP for AIX network module. You rarely need to add or remove any of
those, however, you may want to change the failure detection rate of a
network module.
There are three values to choose from: Fast, Normal and Slow. The normal
heartbeat rate is usually optimal. Speeding up or slowing down failure
detection is an area where you can adjust cluster failover behavior.
If you decide to change the failure detection rate of a network module, keep
the following considerations in mind:
• Failure detection is dependent on the fastest network linking two nodes.
• Faster heartbeat rates may lead to false failure detections, particularly on
busy networks. For example, bursts of high network traffic may delay
heartbeats and this may result in nodes being falsely ejected from the
cluster. Faster heartbeat rates also place a greater load on networks.
• If your networks are very busy and you experience false failure detections,
you can try changing the failure detection speed on the network modules
to slow to avoid this problem.
124
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The failure rate of networks varies, depending on their characteristics. For
example, for an Ethernet, the normal failure detection rate is two keepalives
per second; fast is about four per second; slow is about one per second. For
an HPS network, because no network traffic is allowed when a node joins the
cluster, normal failure detection is 30 seconds; fast is 10 seconds; slow is 60
seconds.
The Change / Show Topology and Group Services Configuration screen
includes the settings for the length of the Topology and Group services logs.
The default settings are highly recommended. The screen also contains
entries for heartbeat settings, but these are not operable (see HACMP/ES
Installation and Administration Guide, SC23-4284, Chapter 18). The
heartbeat rate is now set for each network module in the corresponding
screen (see above).
To learn more about Topology and Group Services, see Chapter 32 of the
HACMP/ES Installation and Administration Guide, SC23-4284.
5.4 NFS considerations
For NFS to work correctly in an HACMP cluster environment, you have to
take care of some special NFS characteristics.
The HACMP scripts have only minimal NFS support. You may need to modify
them to handle your particular configuration. The following sections contain
some suggestions for handling a variety of issues.
5.4.1 Creating Shared Volume Groups
When creating shared volume groups, normally, you can leave the Major
Number field blank and let the system provide a default for you. However,
unless all nodes in your cluster are identically configured, you will have
problems using NFS in an HACMP environment. The reason is that the
system uses the major number as part of the file handle to uniquely identify a
Network File System.
In the event of node failure, NFS clients attached to an HACMP cluster
operate exactly the way they do when a standard NFS server fails and
reboots. If the major numbers are not the same, when another cluster node
takes over the file system and re-exports it, the client application will not
recover, since the file system exported by the node will appear to be different
from the one exported by the failed node.
Cluster Customization
125
Download from Www.Somanuals.com. All Manuals Search And Download.
To prevent problems with NFS file systems in an HACMP cluster, make sure
that each shared volume group has the same major number on all nodes. The
lvlstmajorcommand lists the free major numbers on a node. Use this
command on each node to find a major number that is free on all cluster
nodes, then, record that number in the Major Number field on the Shared
Volume Group/File System (Non-Concurrent Access) worksheet in Appendix
A, Planning Worksheets, of the HACMP for AIX, Version 4.3: Planning Guide,
SC23-4277 for a non-concurrent access configuration.
Alternatively, if you use the Task Guide to create your shared volume groups,
it will make sure that the major number is the same on all nodes that will
share it.
5.4.2 Exporting NFS File Systems
The default scripts provided with HACMP do not use the /etc/exports file.
Instead, the default scripts provided call a cl_export_fsutility that uses the
exportfscommand with the -i flag and specifies the file system names stored
in the HACMP ODM object class.
Therefore export options specified in the /etc/exports file are ignored.
However, export options may be specified by modifying the cl_export_fs
utility. Alternately, the /etc/exports file can be used as is typical in an NFS
environment by simply removing the -iflag from the exportfscommand in
the cl_export_fs utility.
5.4.3 NFS Mounting
For HACMP and NFS to work together properly, you must be aware of the
following mount issues:
5.4.3.1 Creating NFS Mount Points on Clients
A mount point is required in order to mount a file system with NFS. Mount
points are required for NFS clients, not servers; however, you should be
aware that a server can also be a client.
5.4.4 Cascading Takeover with Cross Mounted NFS File Systems
This section describes how to set up cascading resource groups with cross
mounted NFS file systems.
5.4.4.1 Server-to-Server NFS Cross Mounting
HACMP allows you to configure a cluster so that servers can NFS-mount
each other’s file systems. The following figure shows an example:
126
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Figure 14. NFS Cross Mounts
When Node A fails, Node B uses the cl_nfskillutility to close open files in
Node A:/afs, unmounts it, mounts it locally, and re-exports it to waiting clients.
After takeover, Node B has:
/bfs locally mounted
/bfs nfs-exported
/afs locally mounted
/afs nfs-exported
Ensure that the shared volume groups have the same major number on the
server nodes. This allows the clients to re-establish the NFS-mount
transparently after the takeover.
Caveats about Node Names and NFS
In the configuration described above, the node name is used as the NFS
hostname for the mount. This can fail if the node name is not a legitimate
TCP/IP adapter label.
To avoid this problem do one of the following:
Cluster Customization
127
Download from Www.Somanuals.com. All Manuals Search And Download.
• Ensure that node name and the service adapter label are the same on
each node in the cluster
or
• Alias the node name to the service adapter label in the /etc/hosts file.
5.4.5 Cross Mounted NFS File Systems and the Network Lock Manager
If an NFS client application uses the Network Lock Manager, there are
additional considerations to ensure a successful failover. Consider the
following scenario: Node A has a file system mounted locally and exported for
use by clients. Node B is an NFS client and mounts the exported file system
for local use by an application that issues lock requests using the flock()
system call. Node A fails. Node B then attempts to unmount the NFS
mounted file system, mount it as a local file system, and export it for client
use. However, the unmount fails because of outstanding lock requests
against the file system.
Adding the following lines to the cl_deactivate_nfs script will clear
outstanding locks against the failed node and will allow the file system to be
unmounted. However, it will result in the loss of all locks. Consider your
configuration carefully. If you have non-cluster related NFS file systems
where losing locks would be unacceptable, you may need to take appropriate
steps before using this addition to the cl_deactivate_nfs script.
Add the code below between the following two lines (three places):
128
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
######## Add for NFS Lock Removal (start) ########
######## Add for NFS Lock Removal (finish) ########
###############################################################################
#
# Name: cl_deactivate_nfs
#
# Given a list of nfs-mounted filesystems, we try and unmount -f
# any that are currently mounted.
#
# Arguments: list of filesystems.
#
###############################################################################
PROGNAME="$0"
MOUNTED="false"
######## Add for NFS Lock Removal (start) ########
STOPPED="false"
######## Add for NFS Lock Removal (finish) ########
SLEEP="2"
if [ "$VERBOSE_LOGGING" = "high" ]
then
set -x
fi
set -u
if [ $# -ne 0 ]
then
FILELIST=‘for i in $*; do /bin/echo $i; done | /bin/sort -r‘
for fs in $FILELIST
do
# Is the filesystem mounted?
# -s says only return status, -x says exact match
# we use awk instead of cut because mount outputs
# lots of leading blanks that confuse cut
/etc/mount | awk ’{ print $2 }’ | fgrep -s -x "$fs"
if [ $? -eq 0 ]
then
# At least one filesystem is mounted
MOUNTED="true"
# This filesystem is mounted
######## Add for NFS Lock Removal (start) ########
# Determine the host which is making the filesystem
# available
# This will clear any outstanding locks against the
# failed node, not preserve their state, and is thus
# considered a forceful move.
host=‘/etc/mount|grep nfs|grep "$fs"|awk ’{ print $1 }’‘
if [ -n "$host" ]
then
if [ "$STOPPED" = "false" ]
then
stopsrc -s rpc.lockd
stopsrc -s rpc.statd
STOPPED="true"
Cluster Customization
129
Download from Www.Somanuals.com. All Manuals Search And Download.
fi
/bin/rm -f /etc/sm.bak/$host
/bin/rm -f /etc/sm/$host
/bin/rm -f /etc/state
fi
######## Add for NFS Lock Removal (finish) ########
# Send a SIGKILL to all processes having open file
# descriptors within this logical volume to allow
# the unmount to succeed..
cl_nfskill -k -u $fs
fi
done
else
cl_echo 27 ’$PROGNAME: Bad number of arguments ’ $PROGNAME
exit 2
fi
# Make sure all processes have time to die
# Only wait if at least one filesystem is mounted
if [ "$MOUNTED" = "true" ]
then
sleep $SLEEP
fi
FILELIST=‘for i in $*; do /bin/echo $i; done | /bin/sort -r‘
for fs in $FILELIST
do
# Is the filesystem mounted?
# -s says only return status, -x says exact match
# we use awk instead of cut because mount outputs
# lots of leading blanks that confuse cut
/etc/mount | awk ’{ print $2 }’ | fgrep -s -x "$fs"
if [ $? -eq 0 ]
then
# At least one filesystem is mounted
until /etc/umount -f $fs
do
sleep 2
done
fi
done
######## Add for NFS Lock Removal (start) ########
if [ "$STOPPED" = "true" ]
then
startsrc -s rpc.statd
startsrc -s rpc.lockd
fi
######## Add for NFS Lock Removal (finish) ########
exit 0
130
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 6. Cluster Testing
Before you start to test the HACMP configuration, you need to guarantee that
your cluster nodes are in a stable state. Check the state of the:
• Devices
• System parameters
• Processes
• Network adapters
• LVM
• Cluster
• Other items such as SP Switch, printers, and SNA configuration
6.1 Node Verification
Here is a series of suggested actions to test the state of a node before
including HACMP in the testing.
6.1.1 Device State
• Run diag -ain order to clean up the VPD.
• Look in the errorlog for unusual errors by issuing the command errpt |
moreor errpt -a | more.
• Check that all devices are in the available state (lsdev -C | more).
• Check that the SCSI addresses of adapters on shared buses are unique
(lsattr -E -l ascsi0).
• If you are using target mode SCSI networks, check the connection by
issuing cat < /dev/tmscsi#.tmon the first node and cat /etc/hosts >
/dev/tmscsi#.im(enter twice!) on the second node where #is the
appropriate tmscsi device number. Repeat the test in the other direction.
Note that cluster services must be stopped on both nodes to perform this
test.
• To check a serial line between two nodes type stty < /dev/tty#on both
nodes where #is the appropriate tty device number for the RS232
heartbeat connection. Note that cluster services must be stopped on both
nodes to perform this test.
© Copyright IBM Corp. 1999
131
Download from Www.Somanuals.com. All Manuals Search And Download.
6.1.2 System Parameters
• Type dateon all nodes to check that all the nodes in the cluster are
running with their clocks on the same time.
• Ensure that the number of user licenses has been correctly set
(lslicense).
• Check high water mark and other system settings (smitty chgsys).
• Type sysdumpdev -land sysdumpdev -eto ensure that the dump space is
correctly set and that the primary dump device (lslv hd7) is large enough
to accomodate a dump.
• Check that applications to be controlled by HACMP are not started here,
and that extraneous processes which might interfere with HACMP and/or
dominate system resources are not started (more /etc/inittab).
• Check list of cron jobs (crontab -l).
6.1.3 Process State
• Check the paging space usage by issuing lsps -a.
• Look for all expected processes with ps -ef | more.
• Check that the run queue is < 5 and that the CPU usage is at an
acceptable level (vmstat 2 5).
6.1.4 Network State
• Type for example ifconfig lo0, ifconfig en0and ifconfig en1 to check the
network adapter configuration, if you are using ethernet adapters. For
other types of adapters, use the appropriate device name.
• To check the configuration of an SP Switch adapter, type:
/usr/lpp/ssp/css/ifconfig css0.
• Use netstat -ior netstat -into show the network configuration of the
node.
• To check the alternate MAC address, issue netstat -v ent0 | more.
• Look at mbufs sizing relative to requests for memory denied (netstat -m |
more).
• Type netstat -ror netstat -rAnto ensure that there are valid routes to the
other cluster node interfaces and to clients.
• Run no -a | moreand look at the setting of ipforwardingand
ipsendredirects.
132
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Check that all interfaces communicate (ping <ip-address> or ping -R
<ip-address>).
• List the arp table entries with arp -a.
• Check the status of the TCP/IP daemons (lssrc -g tcpip).
• Ensure that there are no bad entries in the /etc/hosts file, especially at the
bottom of the file.
• Verify that, if DNS is in use, the DNS servers are correctly defined (more
/etc/resolv.conf).
• Check the status of NIS by typing ps -ef | grep ypbindand lssrc -g yp.
• The command exportfsshows non-HACMP controlled NFS exports.
• Run snmpinfo -m dump -o /usr/sbin/cluster/hacmp.defs address show
snmp information for Cluster network addresses (including the serial
interfaces).
6.1.5 LVM State
• Ensure that the correct VG's are defined, that quorum and auto-varyon are
correctly defined, and that the shared VG's are in the correct state (lsvg
and lsvg -o).
• Check that there are no stale partitions (lsvg -l).
• Check that all appropriate file systems have been mounted and that none
of the rootvg file systems are full (df -k).
• Check that PVid's have been assigned where necessary and that there
are no ghost disks(lspv).
• Verify that all entries in the /etc/filesystems file are correct and that there
are no erroneous entries (more /etc/filesystemsand lsfs).
6.1.6 Cluster State
• Check the status of the cluster daemons by issuing lssrc -g clusterand
lssrc -g lock.
• Run /usr/sbin/cluster/clstatto check the status of the cluster and the
status of the network interfaces.
• Check the cluster logfiles with tail -f /tmp/hacmp.out, more
/usr/sbin/cluster/history/cluster.mmdd(mmdd = current date), tail -f
/var/adm/cluster.logand more /tmp/cm.log.
• Check that the nodename is correct (odmget HACMPcluster).
Cluster Testing 133
Download from Www.Somanuals.com. All Manuals Search And Download.
• Verify the cluster configuration by running /usr/sbin/cluster/diag/clconfig
-v ’-tr’.
• To show cluster configuration, run: /usr/sbin/cluster/utilities/cllscf.
• To show the clstrmgr version, type: snmpinfo -m dump -o
/usr/sbin/cluster/hacmp.defs clstrmgr.
6.2 Simulate Errors
The following paragraphs will give you hints on how you can simulate
different hardware and software errors in order to verify your HACMP
configuration. As an example, we will use a cluster consisting of two nodes
and a cascading resource group definition.
The term “NodeF” is used for the node to be failed, and the term “NodeT” for
the takeover node of NodeF's resource group.
When executing the test plan, it is helpful to monitor cluster activities during
failover with the following commands. Note that the /tmp/hacmp.out file is the
most useful to monitor, especially if the Debug Level of the HACMP Run Time
Parameters for the nodes has been set to “high”, and if the Application Server
Scripts include the set -xflag and periodic echo commands.
6.2.1 Adapter Failure
The following sections cover adapter failure.
6.2.1.1 Ethernet or Token Ring Interface Failure
In case of an Ethernet or Token Ring interface failure, perform the following
steps:
• Check that all the nodes in the cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor the cluster log files on NodeT.
• Use ifconfig to shut off the appropriate service interface (but not the
Administrative SP Ethernet) on NodeF (for example, ifconfig en0 down).
This will cause the service IP address to failover to the standby adapter on
NodeF.
• Verify that the swap adapter has occurred (including MAC
Addressfailover) and that HACMP has turned the original service interface
back on as the standby interface.
134
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Use ifconfig to swap the service address back to the original service
interface back (ifconfig en1 down). This will cause the service IP address
to failover back to the service adapter on NodeF.
6.2.1.2 Ethernet or Token Ring Adapter or Cable Failure
Perform the following steps in the event of an Ethernet or Token Ring adapter
or cable failure:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor the cluster log files on NodeT.
• Disconnect the network cable from the appropriate service interface (but
not the Administrative SP Ethernet) on NodeF. This will cause the service
IP and MAC addresses to failover to the standby adapter on NodeF.
• Verify that the swap adapter has occurred.
• Reconnect the network cable to the service interface. This will cause the
original service interface to become the standby interface.
• Initiate a swap adapter back to the original service interface by
disconnecting the network cable from the new service interface (originally
the standby interface). This will cause the service IP and MAC addresses
to failover back to the service adapter on NodeF.
• Verify that the swap adapter has occurred.
• Reconnect the cable to the original standby interface.
• Verify that the original standby interface is operating with the standby IP
address.
6.2.1.3 Switch Adapter Failure
Perform the following steps in case of switch adapter failure:
Note
Do not disconnect live switch cables to simulate a switch failure!
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Assign NodeF to be Eprimary.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor the cluster log files on NodeT.
Cluster Testing 135
Download from Www.Somanuals.com. All Manuals Search And Download.
• Generate the switch error in the error log which is being monitored by
HACMP Error Notification (for configuration see 2.6.2.1, “Single
network_down event has been customized, bring down css0 (ifconfig
css0 down) or fence out NodeF from the Control Workstation (Efence NodeF).
• If the first failure simulation method is used, the switch failure will be
detected in the error log (errpt -a | more) on NodeF and cause a node
failover to NodeT. The other two methods will cause HACMP to detect a
network_down event, with the same result. (Note that if there is another
node in the cluster with a lower alphanumeric node name than NodeT,
then that node will become Eprimary. HACMP does not take care of the
Eprimary if a new SP-Switch is used).
• Verify that failover has occurred (netstat -iand pingfor networks, lsvg -o
and viof a test file for volume groups, ps -U <appuid>for application
processes, and Eprimaryfor Eprimary).
• Start HACMP on NodeF (smit clstart). NodeT will release NodeF's
cascading Resource Groups and NodeF will take them back over, but
NodeT (or a lower alphanumeric node) will remain Eprimary.
• Verify that re-integration has occurred (netstat -iand pingfor networks,
lsvg -oand viof a test file for volume groups, ps -U <appuid> for
application processes, and Eprimaryfor Eprimary).
6.2.1.4 Failure of a 7133 Adapter
Perform the following steps in the event of a 7133 Adapter failure:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor cluster logfiles on NodeT if HACMP has been customized to
monitor 7133 disk failures.
• Pull all the cable from the SSA adapter.
• The failure of the 7133 adapter should be detected in the error log (errpt
-a | more) on NodeF or should be noted in the appropriate diagnostics
tool, and the logical volume copies on the disks in drawer 1 will be marked
stale (lsvg -l NodeFvg).
• Verify that all sharedvg file systems and paging spaces are accessible (df
-kand lsps -a).
• Re-attach the cables.
136
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Verify that all sharedvg file systems and paging spaces are accessible (df
-kand lsps -a).
6.2.2 Node Failure / Reintegration
The following sections deal with issues of node failure and reintegration.
6.2.2.1 AIX Crash
Perform the following steps in the event of an AIX crash:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• If NodeF is an SMP, you may want to set the fast reboot switch (mpcfg -cf
11 1).
• Monitor cluster logfiles on NodeT.
• Crash NodeF by entering cat /etc/hosts > /dev/kmem. (The LED on NodeF
will display 888.)
• The OS failure on NodeF will cause a node failover to NodeT.
• Verify that failover has occurred (netstat -iand pingfor networks, lsvg -o
and viof a test file for volume groups, and ps -U <appuid> for application
processes).
• Power cycle NodeF. If HACMP is not configured to start from /etc/inittab,
(on restart) start HACMP on NodeF (smit clstart). NodeF will take back
its cascading Resource Groups.
• Verify that re-integration has occurred (netstat -iand pingfor networks,
lsvg -oand viof a test file for volume groups, and ps -U <appuid> for
application processes).
6.2.2.2 CPU Failure
Perform the following steps in the event of CPU failure:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• If NodeF is an SMP, you may want to set the fast reboot switch (mpcfg -cf
11 1).
• Monitor cluster logfiles on NodeT.
• Power off NodeF. This will cause a node failover to NodeT.
Cluster Testing 137
Download from Www.Somanuals.com. All Manuals Search And Download.
• Verify that failover has occurred (netstat -iand pingfor networks, lsvg -o
and viof a test file for volume groups, and ps -U <appuid> for application
processes).
• Power cycle NodeF. If HACMP is not configured to start from /etc/inittab
(on restart), start HACMP on NodeF (smit clstart). NodeF will take back
its cascading Resource Groups.
• Verify that re-integration has occurred (netstat -iand pingfor networks,
lsvg -oand viof a test file for volume groups, and ps -U <appuid> for
application processes).
6.2.2.3 TCP/IP Subsystem Failure
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor the cluster log files on NodeT.
• On NodeF, stop the TCP/IP subsystem (sh /etc/tcp.clean) or crash the
subsystem by increasing the size of the sb_max and thewall parameters to
large values (no -o sb_max=10000; no -o thewall=10000) and ping NodeT.
Note that you should record the values for sb_max and thewall prior to
modifying them, and, as an extra check, you may want to add the original
values to the end of /etc/rc.net.
• The TCP/IP subsystem failure on NodeF will cause a network failure of all
the TCP/IP networks on NodeF. Unless there has been some
customization done to promote this type of failure to a node failure, only
the network failure will occur. The presence of a non-TCP/IP network
(RS232, target mode SCSI or target mode SSA) should prevent the cluster
from triggering a node down in this situation.
• Verify that the network_down event has been run by checking the
/tmp/hacmp.out file on either node. By default, the network_down script
does nothing, but it can be customized to do whatever is appropriate for
that situation in your environment.
• On NodeF, issue the command startsrc -g tcpip. This should restart the
TCP/IP daemons, and should cause a network_up event to be triggered in
the cluster for each of your TCP/IP networks.
6.2.3 Network Failure
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
138
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Monitor the cluster log files on NodeT.
• Disconnect the network cable from the appropriate service and all the
standby interfaces at the same time (but not the Administrative SP
Ethernet) on NodeF. This will cause HACMP to detect a network_down
event.
• HACMP triggers events dependent on your configuration of the
network_down event. By default, no action is triggered by the
network_down event.
• Verify that the expected action has occurred.
6.2.4 Disk Failure
The following sections deal with issues of disk failure.
6.2.4.1 Mirrored rootvg Disk (hdisk0) Failure
Perform the following steps in case of mirrored rootvg disk (hdisk0) failure:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Verify that the bootlist contains hdisk0 and hdisk1, if for example, hdisk1 is
the mirror of hdisk0 (bootlist -m normal -o).
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor cluster logfiles on NodeT if HACMP has been customized to
monitor SCSI disk failures.
• Slide back cover/casing on NodeF to get access to hdisk0 (this may first
require turning the key to service mode). Pull the power cable (several
colored wires with a white plastic connector) from the rear of hdisk0 (the
lower internal disk is hdisk0, and the upper internal disk is hdisk1 on most
systems). If you have a hot-pluggable disk, just pull the disk out of the
frame.
• The failure of hdisk0 should be detected in the error log (errpt -a | more)
on NodeF.
• Verify that all rootvg file systems and paging spaces are accessible (df;
lsps -a).
• Shutdown (smit clstop; shutdown -F) and power off NodeF.
• Turn key to normal mode, power on NodeF, and verify that the system
boots correctly. Log in and verify that all the rootvg file systems have been
mounted (df).
• Shutdown (shutdown -F) and power off NodeF.
Cluster Testing 139
Download from Www.Somanuals.com. All Manuals Search And Download.
• Reconnect hdisk0, close the casing, and turn the key to normal mode.
• Power on NodeF then verify that the rootvg logical volumes are no longer
stale (lsvg -l rootvg).
6.2.4.2 7135 Disk Failure
Perform the following steps in the event of a disk failure:
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
• Monitor cluster logfiles on NodeT if HACMP has been customized to
monitor 7135 disk failures.
• Mark a shared disk failed through smit (smit raidiant; RAIDiant Disk
Array Manager -> Change/Show Drive Status -> select the appropriate
hdisk -> select the appropriate physical disk -> F4 to select a Drive
Status of 83 Fail Drive), or if the disk is hot pluggable, remove the disk.
• The amber light on the front of the 7135 comes on, and can also be seen
in SMIT (smit raidiant; RAIDiant Disk Array Manager -> List all SCSI
RAID Arrays).
• Verify that all sharedvg file systems and paging spaces are accessible (df
and lsps -a).
• If using RAID5 with Hot Spare, verify that reconstruction has completed to
the Hot Spare, then un-mark or plug the failed disk back in. If using
RAID1, sync the volume group (syncvg NodeFvg).
• If using RAID5 without Hot Spare, mark the failed disk Optimal (smit
raidiant; RAIDiant Disk Array Manager -> Change/Show Drive Status;
select the appropriate hdisk -> select the appropriate physical disk ->
F4 to select a Drive Status of 84 Replace and Reconstruct Drive).
• Verify that the reconstruction has completed (smit raidiant;RAIDiant Disk
Array Manager -> List all SCSI RAID Arrays).
• Verify that all sharedvg file systems and paging spaces are accessible (df
and lsps -a) and that the partitions are not stale (lsvg -l sharedvg). Also
verify that the yellow light has turned off on the 7135.
6.2.4.3 Mirrored 7133 Disk Failure
• Check, by way of the verification commands, that all the Nodes in the
cluster are up and running.
• Optional: Prune the error log on NodeF (errclear 0).
140
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• Monitor cluster logfiles on NodeT if HACMP has been customized to
monitor 7133 disk failures.
• Since the 7133 disk is hot pluggable, remove a disk from drawer 1
associated with NodeF's shared volume group.
• The failure of the 7133 disk will be detected in the error log (errpt -a |
more) on NodeF, and the logical volumes with copies on that disk will be
marked stale (lsvg -l NodeFvg).
• Verify that all NodeFvg file systems and paging spaces are accessible (df
-k and lsps -a).
• Plug the failed disk back in, then sync the volume group (syncvg NodeFvg).
• Verify that all NodeFvg file systems and paging spaces are accessible (df
-kand lsps -a) and that the partitions are not stale (lsvg -l NodeFvg).
6.2.5 Application Failure
By default, HACMP does not recognize application failures. With some
additional configuration, it is possible to “teach” HACMP application failures
and trigger events (for more information see 2.6.2.3, “Application Failure” on
page 47).
So, the way of testing application failures is strongly dependent on your
configuration.
Before you start to do the configuration and the testing application failure
notification, analyse your application for possible failures. Then try to
reproduce them.
Cluster Testing 141
Download from Www.Somanuals.com. All Manuals Search And Download.
142
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 7. Cluster Troubleshooting
Typically, a functioning HACMP cluster requires minimal intervention. If a
problem occurs, however, diagnostic and recovery skills are essential. Thus,
troubleshooting requires that you identify the problem quickly and apply your
understanding of the HACMP for AIX software to restore the cluster to full
operation.
In general, troubleshooting an HACMP cluster involves:
• Becoming aware that a problem exists
• Determining the source of the problem
• Correcting the problem
Becoming aware of a problem is often through system messages on the
console, end-users complaining about slow or unavailable services or
through some sort of monitoring of your cluster. When an HACMP for AIX
script or daemon generates a message, the message is written to the system
console and to one or more cluster log files. Messages written to the system
console may scroll off screen before you notice them. The following
paragraphs provide an overview of the log files, which are to be consulted for
cluster troubleshooting, as well as some information on specific cluster states
you may find there.
7.1 Cluster Log Files
HACMP for AIX scripts, daemons, and utilities write messages to the
following log files:
Table 21. HACMP Log Files
Log File Name
Description
/var/adm/cluster.log
Contains time-stamped, formatted messages generated by
HACMP for AIX scripts and daemons. In this log file, there
is one line written for the start of each event, and one line
written for the completion.
/tmp/hacmp.out
Contains time-stamped, formatted messages generated by
the HACMP for AIX scripts. In verbose mode, this log file
contains a line-by-line record of each command executed
in the scripts, including the values of the arguments passed
to the commands. By default, the HACMP for AIX software
writes verbose information to this log file; however, you can
change this default. Verbose mode is recommended.
© Copyright IBM Corp. 1999
143
Download from Www.Somanuals.com. All Manuals Search And Download.
Log File Name
Description
system error log
Contains time-stamped, formatted messages from all AIX
subsystems, including the HACMP for AIX scripts and
daemons.
/usr/sbin/cluster/
history/cluster.mmdd
Contains time-stamped, formatted messages generated by
the HACMP for AIX scripts. The system creates a new
cluster history log file every day that has a cluster event
occurring. It identifies each day’s file by the filename
extension, where mm indicates the month and dd indicates
the day.
/tmp/cm.log
Contains time-stamped, formatted messages generated by
HACMP for AIX clstrmgr activity. Information in this file is
used by IBM Support personnel when the clstrmgr is in
debug mode. Note that this file is overwritten every time
cluster services are started; so, you should be careful to
make a copy of it before restarting cluster services on a
failed node.
/tmp/cspoc.log
Contains time-stamped, formatted messages generated by
HACMP for AIX C-SPOC commands. Because the
C-SPOC utility lets you start or stop the cluster from a
single cluster node, the /tmp/cspoc.log is stored on the
node that initiates a C-SPOC command.
/tmp/dms_logs.out
/tmp/emuhacmp.out
Stores log messages every time HACMP for AIX triggers
the deadman switch.
Contains time-stamped, formatted messages generated by
the HACMP for AIX Event Emulator. The messages are
collected from output files on each node of the cluster, and
cataloged together into the /tmp/emuhacmp.outlog file.In
verbose mode (recommended), this log file contains a
line-by-line record of every event emulated. Customized
scripts within the event are displayed, but commands
within those scripts are not executed.
For a more detailed description of the cluster log files consult Chapter 2 of the
HACMP for AIX, Version 4.3: Troubleshooting Guide, SC23-4280.
7.2 config_too_long
If the cluster manager recognizes a state change in the cluster, it acts upon it
by executing an event script. However, some circumstances, like errors within
the script or special conditions of the cluster, might cause the event script to
144
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
hang. After a certain amount of time, by default 360 seconds, the cluster
manager will issue a config_too_long message into the /tmp/hacmp.out file.
The message issued looks like this:
The cluster has been in reconfiguration too long;Something may be wrong.
In most cases, this is because an event script has failed. You can find out
more by analyzing the /tmp/hacmp.out file.The error messages in the
/var/adm/cluster.log file may also be helpful. You can then fix the problem
identified in the log file and execute the clruncmdcommand on the command
line, or by using the SMIT Cluster Recovery Aids screen. The clruncmd
command signals the Cluster Manager to resume cluster processing.
Note, however, that sometimes scripts simply take too long, so the message
showing up isn’t always an error, but sometimes a warning. If the message is
issued, that doesn’t necessarily mean that the script failed or never finished.
A script running for more than 360 seconds can still be working on something
and eventually get the job done. Therefore, it is essential to look at the
/tmp/hacmp.out file to find out what is actually happening.
7.3 Deadman Switch
The term “deadman switch” describes the AIX kernel extension that causes a
system panic and dump under certain cluster conditions if it is not reset. The
deadman switch halts a node when it enters a hung state that extends
beyond a certain time limit. This enables another node in the cluster to
acquire the hung node’s resources in an orderly fashion, avoiding possible
contention problems.
If this is happening, and it isn’t obvious why the cluster manager was kept
from resetting this timer counter, for example because some application ran
at a higher priority as the clstrmgrprocess, customizations related to
performance problems should be performed in the following order:
1. Tune the system using I/O pacing.
2. Increase the syncdfrequency.
3. If needed, increase the amount of memory available for the
communications subsystem.
4. Change the Failure Detection Rate.
Each of these options is described in the following sections.
Cluster Troubleshooting 145
Download from Www.Somanuals.com. All Manuals Search And Download.
7.3.1 Tuning the System Using I/O Pacing
Use I/O pacing to tune the system so that system resources are distributed
more equitably during large disk writes. Enabling I/O pacing is required for an
HACMP cluster to behave correctly during large disk writes, and it is strongly
recommended if you anticipate large blocks of disk writes on your HACMP
cluster.
You can enable I/O pacing using the smit chgsysfastpath to set high- and
low-water marks. These marks are by default set to zero (disabling I/O
pacing) when AIX is installed. While the most efficient high- and low-water
marks vary from system to system, an initial high-water mark of 33 and a
low-water mark of 24 provide a good starting point. These settings only
slightly reduce write times, and consistently generate correct failover
behavior from HACMP for AIX. If a process tries to write to a file at the
high-water mark, it must wait until enough I/O operations have finished to
make the low-water mark. See the AIX Performance Monitoring & Tuning
Guide, SC23-2365 for more information on I/O pacing.
7.3.2 Extending the syncd Frequency
Edit the /sbin/rc.boot file to increase the syncdfrequency from its default
value of 60 seconds to either 30, 20, or 10 seconds. Increasing the frequency
forces more frequent I/O flushes and reduces the likelihood of triggering the
deadman switch due to heavy I/O traffic.
7.3.3 Increase Amount of Memory for Communications Subsystem
If the output of netstat -m reports that requests for mbufs are being denied,
or if errors indicating LOW_MBUFSare being logged to the AIX error report,
increase the value associated with “thewall” network option. The default
value is 25% of the real memory. This can be increased to as much as 50% of
the real memory.
To change this value, add a line similar to the following at the end of the
/etc/rc.net file:
no -o thewall=xxxxx
where xxxxx is the value you want to be available for use by the
communications subsystem. For example,
no -o thewall=65536
146
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
7.3.4 Changing the Failure Detection Rate
Use the SMIT Change/Show a Cluster Network Modulescreen to change the
failure detection rate for your network module only if enabling I/O pacing or
extending the syncd frequency did not resolve deadman problems in your
cluster. By changing the failure detection rate to “Slow”, you can extend the
time required before the deadman switch is invoked on a hung node and
before a takeover node detects a node failure and acquires a hung node’s
resources. See the HACMP for AIX, Version 4.3: Administration Guide,
SC23-4279 for more information and instructions on changing the Failure
Detection Rate.
Note
I/O pacing must be enabled before completing these procedures; it
regulates the number of I/O data transfers. Also, keep in mind that the Slow
setting for the Failure Detection Rate is network specific.
7.4 Node Isolation and Partitioned Clusters
Node isolation occurs when all networks connecting nodes fail but the nodes
remain up and running. One or more nodes can then be completely isolated
from the others. A cluster in which this has happened is called a partitioned
cluster. A partitioned cluster has two groups of nodes (one or more in each),
neither of which cannot communicate with the other. Let’s consider a two
node cluster where all networks have failed between the two nodes, but each
node remains up and running.
The problem with a partitioned cluster is that each node interprets the
absence of keepalives from its partner to mean that the other node has failed,
and then generates node failure events. Once this occurs, each node
attempts to take over resources from a node that is still active and therefore
still legitimately owns those resources. These attempted takeovers can cause
unpredictable results in the cluster—for example, data corruption due to a
disk being reset.
To guard against a TCP/IP subsystem failure causing node isolation, the
nodes should also be connected by a point-to-point serial network. This
connection reduces the chance of node isolation by allowing the Cluster
Managers to communicate even when all TCP/IP-based networks fail.
It is important to understand that the serial network does not carry TCP/IP
communication between nodes; it only allows nodes to exchange keepalives
Cluster Troubleshooting 147
Download from Www.Somanuals.com. All Manuals Search And Download.
and control messages so that the Cluster Manager has accurate information
about the status of its partner.
When a cluster becomes partitioned, and the network problem is cleared after
the point when takeover processing has begun so that keepalive packets
start flowing between the partitioned nodes again, something must be done to
restore order in the cluster. This order is restored by the DGSP Message.
7.5 The DGSP Message
A DGSP message (short for Diagnostic Group Shutdown Partition) is sent
when a node loses communication with the cluster and then tries to
re-establish communication.
For example, if a cluster node becomes unable to communicate with other
nodes, yet it continues to work through its process table, the other nodes
conclude that the “missing” node has failed because they no longer are
receiving keepalive messages from it. The remaining nodes then process the
necessary events to acquire the disks, IP addresses, and other resources
from the “missing” node. This attempt to take over resources results in the
dual-attached disks receiving resets to release them from the “missing” node
and the start of IP address takeover scripts.
As the disks are being acquired by the takeover node (or after the disks have
been acquired and applications are running), the “missing” node completes
its process table (or clears an application problem) and attempts to resend
keepalive messages and rejoin the cluster. Since the disks and IP addresses
are in the process of being successfully taken over, it becomes possible to
have a duplicate IP address on the network and the disks may start to
experience extraneous traffic on the data bus.
Because the reason for the “missing” node remains undetermined, you can
assume that the problem may repeat itself later, causing additional down time
of not only the node but also the cluster and its applications. Thus, to ensure
the highest cluster availability, a DGSP message is sent to all nodes in one of
the partitions. Any node receiving a DGSP message halts immediately, in
order to not cause any damage on disks or confusion on the networks.
In a partitioned cluster situation, the smaller partition (lesser number of
nodes) is shut down, with each of its nodes getting a DGSP message. If the
partitions are of equal size, the one with the node name beginning in the
lowest name in the alphabet gets shut down. For example, in a cluster where
one partition has NodeA and the other has NodeB, NodeB will be shut down.
148
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
7.6 User ID Problems
Within an HACMP cluster, you always have more than one node potentially
offering the same service to a specific user or a specific user id.
As the node providing the service can change, the system administrator has
to ensure that the same user and group is known to all nodes potentially
running an application. So, in case one node is failing, and the application is
taken over by the standby node, a user can go on working since the takeover
node knows that user under exactly the same user and group id.
Since user access within an NFS mounted file system is granted based on
user IDs, the same applies to NFS mounted file systems.
For more information on managing user and group accounts within a cluster,
“Managing User and Groups in a Cluster” of the HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279.
7.7 Troubleshooting Strategy
In order to quickly find a solution to a problem in the cluster, some sort of
strategy is helpful for pinpointing the problem. The following guidelines
should make the troubleshooting process more productive:
• Save the log files associated with the problem before they become
unavailable. Make sure you save the /tmp/hacmp.out and /tmp/cm.log files
before you do anything else to try to figure out the cause of the problem.
• Attempt to duplicate the problem. Do not rely too heavily on the user’s
problem report. The user has only seen the problem from the application
level. If necessary, obtain the user’s data files to recreate the problem.
• Approach the problem methodically. Allow the information gathered from
each test to guide your next test. Do not jump back and forth between
tests based on hunches.
• Keep an open mind. Do not assume too much about the source of the
problem. Test each possibility and base your conclusions on the evidence
of the tests.
• Isolate the problem. When tracking down a problem within an HACMP
cluster, isolate each component of the system that can fail and determine
whether it is working. Work from top to bottom, following the progression
described in the following section.
Cluster Troubleshooting 149
Download from Www.Somanuals.com. All Manuals Search And Download.
• Go from the simple to the complex. Make the simple tests first. Do not try
anything complex and complicated until you have ruled out the simple and
obvious.
• Do not make more than one change at a time. If you do, and one of the
changes corrects the problem, you have no way of knowing which change
actually fixed the problem. Make one change, test the change, and then, if
necessary, make the next change.
• Do not neglect the obvious. Small things can cause big problems. Check
plugs, connectors, cables, and so on.
• Keep a record of the tests you have completed. Record your tests and
results, and keep an historical record of the problem, in case it reappears.
150
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 8. Cluster Management and Administration
This chapter covers all aspects of monitoring and managing an existing
HACMP cluster. This includes a description of the different monitoring
methods and tools available, how to start and stop the cluster, changing
cluster or resource configurations, applying software fixes, user
management, and other things.
8.1 Monitoring the Cluster
By design, HACMP for AIX compensates for various failures that occur within
a cluster. For example, HACMP for AIX compensates for a network adapter
failure by swapping in a standby adapter. As a result, it is possible that a
component in the cluster could have failed and that you would be unaware of
the fact. The danger here is that, while HACMP for AIX can survive one or
possibly several failures, a failure that escapes your notice threatens a
cluster’s ability to maintain a highly available environment.
HACMP for AIX provides the following tools for monitoring an HACMP cluster:
• The /usr/sbin/cluster/clstat utility, which reports the status of key
cluster components—the cluster itself, the nodes in the cluster, and the
network adapters connected to the nodes.
• The HAViewutility, which monitors HACMP clusters through the NetView for
AIX graphical network management interface. It lets users monitor
multiple HACMP clusters and cluster components across a network from a
single node.
• The SMIT Show Cluster Servicesscreen, which shows the status of the
HACMP for AIX daemons
• The following log files: the /var/adm/cluster.log file, which tracks cluster
events, the /tmp/hacmp.out file, which records the output generated by
configuration scripts as they execute, the
/usr/sbin/cluster/history/cluster.mmdd log file, which logs the daily cluster
history, and the /tmp/cspoc.log file, which logs the status of C-SPOC
commands executed on cluster nodes.
When you monitor a cluster, use the clstatutility to examine the cluster and
its components. Also, constantly monitor the /tmp/hacmp.out file. Use the
SMIT Show Cluster Services screen to make sure that the necessary HACMP
for AIX daemons are running on each node. Finally, if necessary, examine the
other cluster log files to get a more in-depth view of the cluster status.
© Copyright IBM Corp. 1999
151
Download from Www.Somanuals.com. All Manuals Search And Download.
Consult the HACMP for AIX, Version 4.3: Troubleshooting Guide, SC23-4280,
for help if you detect a problem with an HACMP cluster.
8.1.1 The clstat Command
HACMP for AIX provides the /usr/sbin/cluster/clstatcommand for
monitoring a cluster and its components. The clstatutility is a clinfo client
program that uses the Clinfo API to retrieve information about the cluster.
Clinfo must be running on a node for this utility to work properly.
The /usr/sbin/cluster/clstatutility runs on both ASCII and X Window
Display clients in either single-cluster or multi-cluster mode. Multi-cluster
mode requires that you use the -iflag when invoking the clstatutility. The
client display automatically corresponds to the capability of the system. For
example, if you run clstaton an X Window client, a graphical display for the
utility appears. However, you can run an ASCII display on an X-capable
machine by specifying the -a flag. In order to set up a connection to the
cluster nodes, the /usr/sbin/cluster/etc/clhosts file must be configured on the
client.
The clstatutility reports whether the cluster is up, down, or unstable. It also
reports whether a node is up, down, joining, leaving, or reconfiguring, and the
number of nodes in the cluster. For each node, the utility displays the IP label
and address of each network interface attached to the node, and whether that
interface is up or down. See the clstatman page for additional information
about this utility.
8.1.2 Monitoring Clusters using HAView
HAView is a cluster monitoring utility that allows you to monitor HACMP
clusters using NetView for AIX. Using NetView, you can monitor clusters and
cluster components across a network from a single management station.
HAView creates and modifies NetView objects that represent clusters and
cluster components. It also creates submaps that present information about
the state of all nodes, networks, and network interfaces associated with a
particular cluster. This cluster status and configuration information is
accessible through NetView’s menu bar.
HAView monitors cluster status using the Simple Network Management
Protocol (SNMP). It combines periodic polling and event notification through
traps to retrieve cluster topology and state changes from the HACMP
management agent, that is, the Cluster SMUX peer daemon (clsmuxpd).
152
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
More details on how to configure HAView and on how to monitor your cluster
with HAView can be found in Chapter 3, “Monitoring an HACMP cluster” in
HACMP for AIX, Version 4.3: Administration Guide, SC23-4279.
8.1.3 Cluster Log Files
HACMP for AIX writes the messages it generates to the system console and
to several log files. Because each log file contains a different subset of the
types of messages generated by HACMP for AIX, you can get different views
of cluster status by viewing different log files. HACMP for AIX writes
messages into the log files described below. See Chapter 2, “Examining
Cluster Log Files”, in the HACMP for AIX, Version 4.3: Troubleshooting
Guide, SC23-4280 for more information about these files.
8.1.3.1 /var/adm/cluster.log
The /var/adm/cluster.log file is the main HACMP for AIX log file. HACMP
error messages and messages about HACMP for AIX-related events are
appended to this log with the time and date when they occurred.
8.1.3.2 /tmp/hacmp.out
The /tmp/hacmp.out file records the output generated by the configuration
and startup scripts as they execute. This information supplements and
expands upon the information in the /var/adm/cluster.log file. To receive
verbose output, the Debug Level run-time parameter should be set to high,
which is the default.
8.1.3.3 /usr/sbin/cluster/history/cluster.mmdd
The /usr/sbin/cluster/history/cluster.mmdd file contains timestamped,
formatted messages generated by HACMP for AIX scripts. The system
creates a cluster history file whenever cluster events occur, identifying each
file by the file name extension mmdd, where mm indicates the month and dd
indicates the day.
While it is more likely that you will use these files during troubleshooting, you
should occasionally look at them to get a more detailed picture of the activity
within a cluster.
8.1.3.4 System Error Log
The system error log file contains timestamped, formatted messages from all
AIX subsystems, including HACMP for AIX scripts and daemons. Cluster
events are logged as operator messages (error id: AA8AB241) in the system
error log.
Cluster Management and Administration 153
Download from Www.Somanuals.com. All Manuals Search And Download.
8.1.3.5 /tmp/cm.log
Contains timestamped, formatted messages generated by HACMP for AIX
clstrmgractivity. This file is typically used by IBM support personnel.
8.1.3.6 /tmp/cspoc.log
Contains timestamped, formatted messages generated by HACMP for AIX
C-SPOC commands. The /tmp/cspoc.log file resides on the node that invokes
the C-SPOC command.
8.1.3.7 /tmp/emuhacmp.out
The /tmp/emuhacmp.out file records the output generated by the event
emulator scripts as they execute. The /tmp/emuhacmp.out file resides on the
node from which the event emulator is invoked. You can use the environment
variable EMUL_OUTPUT to specify another name and location for this file,
but the format and information remains the same.
With HACMP/ES, because of its RSCT technology, there are 3 more logfiles
you may want to watch. These are:
8.1.3.8 /var/ha/log/grpsvcs.<filename>
Contains timestamped messages in ASCII format. These track the execution
of internal activities of the grpsvcs daemon. IBM support personnel use this
information for troubleshooting. The file gets trimmed regularly. Therefore,
please save it promptly if there is a chance you may need it.
8.1.3.9 /var/ha/log/topsvcs.<filename>
Contains timestamped messages in ASCII format. These track the execution
of internal activities of the topsvcs daemon. IBM support personnel use this
information for troubleshooting. The file gets trimmed regularly. Therefore,
please save it promptly if there is a chance you may need it.
8.1.3.10 /var/ha/log/grpglsm
The /var/ha/log/grpglsm file tracks the execution of internal activities of the
grpglsm daemon. IBM support personnel use this information for
troubleshooting. The file gets trimmed regularly. Therefore please save it
promptly if there is a chance you may need it.
8.2 Starting and Stopping HACMP on a Node or a Client
This paragraph explains how to start and stop cluster services on cluster
nodes and clients. It also describes how the Cluster-Single Point of Control
154
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
(C-SPOC) utility can be used to start and stop cluster services on all nodes in
cluster environments.
Starting cluster services refers to the process of starting the HACMP for AIX
daemons that enable the coordination required between nodes in a cluster.
Starting cluster services on a node also triggers the execution of certain
HACMP for AIX scripts that initiate the cluster. Stopping cluster services
refers to stopping these same daemons on a node. This action may or may
not cause the execution of additional HACMP for AIX scripts, depending on
the type of shutdown you perform.
8.2.1 HACMP Daemons
The following lists the required and optional HACMP for AIX daemons.
8.2.1.1 Cluster Manager daemon (clstrmgr)
This daemon maintains the heartbeat protocol between the nodes in the
cluster, monitors the status of the nodes and their interfaces, and invokes the
appropriate scripts in response to node or network events. All cluster nodes
must run the clstrmgrdaemon.
8.2.1.2 Cluster SMUX Peer daemon (clsmuxpd)
This daemon maintains status information about cluster objects. This daemon
works in conjunction with the Simple Network Management Protocol (snmpd)
daemon. All cluster nodes must run the clsmuxpddaemon.
Note
The clsmuxpddaemon cannot be started unless the snmpddaemon is
running.
8.2.1.3 Cluster Lock Manager daemon (cllockd)
This daemon provides advisory locking services. The cllockddaemon may be
required on cluster nodes if those nodes are part of a concurrent access
configuration, but this is not necessarily so. Check with your application
vendor to see if it is required.
Note
If the clsmuxpddaemon or the cllockddaemon cannot be started by the
Cluster Manager (e. g. the ports are already in use), the Cluster Manager
logs an error message and dies.
Cluster Management and Administration 155
Download from Www.Somanuals.com. All Manuals Search And Download.
8.2.1.4 Cluster Information Program daemon (clinfo)
This daemon provides status information about the cluster to cluster nodes
and clients and invokes the /usr/sbin/cluster/etc/clinfo.rcscript in
response to a cluster event. The clinfodaemon is optional on cluster nodes
and clients. However, it is a prerequisite for running the clstat utility.
With RSCT (RISC System Cluster Technology) on HACMP/ES Version 4.3,
there are several more daemons.
8.2.1.5 Cluster Topology Services daemon (topsvcsd)
This daemon monitors the status of network adapters in the cluster. All
HACMP/ES cluster nodes must run the topsvcsddaemon
8.2.1.6 Cluster Event Management daemon (emsvcsd)
This daemon matches information about the state of system resources with
information about resource conditions of interest to client programs
(applications, subsystems, and other programs).The emsvcsddaemon runs on
each node of a domain.
8.2.1.7 Cluster Group Services daemon (grpsvcsd)
This daemon manages all of the distributed protocols required for cluster
operation. All HACMP/ES cluster nodes must run the grpsvcsddaemon.
8.2.1.8 Cluster Globalized Server Daemon daemon (grpglsmd)
This daemon operates as a grpsvcsclient; its function is to make switch
adapter membership global across all cluster nodes. All HACMP/ES cluster
nodes must run the grpglsmddaemon.
8.2.2 Starting Cluster Services on a Node
You start cluster services on a node by executing the HACMP
/usr/sbin/cluster/etc/rc.clusterscript. Use the Start Cluster Services SMIT
screen to build and execute this command. The rc.cluster script initializes
the environment required for HACMP by setting environment variables and
then calls the /usr/sbin/cluster/utilities/clstart script to start the HACMP
daemons. The clstart script is the HACMP script that starts all the cluster
services. It does this by calling the SRC startsrccommand to start the
specified subsystem or group.
Using the C-SPOC utility, you can start cluster services on any node (or on all
nodes) in a cluster by executing the C-SPOC
/usr/sbin/cluster/utilities/cl_rc.clustercommand on a single cluster
node. The C-SPOC cl_rc.clustercommand calls the rc.clustercommand to
start cluster services on the nodes specified from the one node. The nodes
156
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
are started in sequential order - not in parallel. The output of the command
run on the remote node is returned to the originating node. Because the
command is executed remotely, there can be a delay before the command
output is returned.
8.2.2.1 Automatically Restarting Cluster Services
You can optionally have cluster services start whenever the system is
rebooted. If you specify the -Rflag to the rc.cluster command, or specify
restart or both in the Start Cluster Services SMIT screen, the rc.cluster
script adds the following line to the /etc/inittab file.
hacmp:2:wait:/usr/sbin/cluster/etc/rc.cluster -boot> /dev/console 2>&1
# Bring up Cluster
At system boot, this entry causes AIX to execute the
/usr/sbin/cluster/etc/rc.clusterscript to start HACMP Cluster Services.
Note
Be aware that if the cluster services are set to restart automatically at boot
time, you may face problems with node integration after a power failure and
restoration, or you may want to test a node after doing maintenance work
before having it rejoin the cluster.
8.2.2.2 Starting Cluster Services with IP Address Takeover Enabled
If IP address takeover is enabled, the /usr/sbin/cluster/etc/rc.clusterscript
calls the /etc/rc.net script to configure and start the TCP/IP interfaces and to
set the required network options.
8.2.3 Stopping Cluster Services on a Node
You stop cluster services on a node by executing the HACMP
/usr/sbin/cluster/etc/clstopscript. Use the HACMP for AIX Stop Cluster
Services SMIT screen to build and execute this command. The clstopscript
stops an HACMP daemon or daemons. The clstopscript starts all the cluster
services or individual cluster services by calling the SRC command stopsrc.
Using the C-SPOC utility, you can stop cluster services on a single node or on
all nodes in a cluster by executing the C-SPOC
/usr/sbin/cluster/utilities/cl_clstopcommand on a single node. The
C-SPOC cl_clstopcommand performs some cluster-wide verification and
then calls the clstopcommand to stop cluster services on the specified
nodes. The nodes are stopped in sequential order—not in parallel. The output
of the command that is run on the remote node is returned to the originating
Cluster Management and Administration 157
Download from Www.Somanuals.com. All Manuals Search And Download.
node. Because the command is executed remotely, there can be a delay
before the command output is returned.
8.2.3.1 When to Stop Cluster services
You typically stop cluster services in the following situations:
• Before making any hardware or software changes or other scheduled
node shutdowns or reboots. Failing to do so may cause unintended cluster
events to be triggered on other nodes.
• Before certain reconfiguration activity. Some changes to the cluster
information stored in the ODM require stopping and restarting the cluster
services on all nodes for the changes to become active. For example, if
you wish to change the name of the cluster, the name of a node, or the
name of an adapter, you must stop and restart the cluster.
8.2.3.2 Types of Cluster Stops
When you stop cluster services, you must also decide how to handle the
resources that were owned by the node you are removing from the cluster.
You have the following options:
Graceful
In a graceful stop, the HACMP software shuts
down its applications and releases its resources.
The other nodes do not take over the resources of
the stopped node.
Graceful with Takeover In a graceful with takeover stop, the HACMP
software shuts down its applications and releases
its resources. The surviving nodes take over these
resources. This is also called intentional failover.
Forced
In a forced stop, the HACMP daemons only are
stopped, without releasing any resources. For
example, the stopped node stays on its service
address if IP Address Takeover has been enabled.
It does not stop its applications, unmount its file
systems or varyoff its shared volume groups. The
other nodes do not take over the resources of the
stopped node. Please note that the forced option
is currently not supported at the Version 4.3 level
in HACMP/ES, only in HACMP Classic.
8.2.3.3 Abnormal Termination of a Cluster Daemon
If the SRC detects that any HACMP daemon has exited abnormally (without
being shut down using the clstopcommand), it executes the
/usr/sbin/cluster/utilities/clexit.rc script to halt the system. This
158
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
prevents unpredictable behavior from corrupting the data on the shared
disks. See the clexit.rcman page for additional information.
Important Note
Never use the kill -9command on the clstrmgrdaemon. Using the kill
command causes the clstrmgrdaemon to exit abnormally. This causes the
SRC to run the /usr/sbin/cluster/utilities/clexit.rc script which halts the
system immediately, causing the surviving nodes to initiate failover.
8.2.4 Starting and Stopping Cluster Services on Clients
Use the /usr/sbin/cluster/etc/rc.cluster script or the startsrccommand to
start clinfoon a client, as shown below:
/usr/sbin/cluster/etc/rc.cluster
You can also use the standard AIX startsrccommand:
startsrc -s clinfo
Use the standard AIX stopsrccommand to stop clinfoon a client machine:
stopsrc -s clinfo
8.2.4.1 Maintaining Cluster Information Services on Clients
In order for the clinfo daemon to get the information it needs, you must edit
the /usr/sbin/cluster/etc/clhosts file. As installed, the clhosts file on an
HACMP client node contains no hostnames or addresses. HACMP server
addresses must be provided by the user at installation time. This file should
contain all boot and service names or addresses of HACMP servers from any
cluster accessible through logical connections with this client node. Upon
startup, clinfouses these names or addresses to attempt communication
with a clsmuxpdprocess executing on an HACMP server.
An example list of hostnames/addresses in a clhosts file follows:
n0_cl83 # n0 service
n2_cl83 # n2 service
n3_cl83 # n3 service
For more detailed information on the clinfo command refer to Chapter 2,
“Starting and Stopping Cluster Services”, HACMP for AIX, Version 4.3:
Administration Guide, SC23-4279.
Cluster Management and Administration 159
Download from Www.Somanuals.com. All Manuals Search And Download.
8.3 Replacing Failed Components
From time to time, it will be necessary to perform hardware maintenance or
upgrades on cluster components. Some replacements or upgrades can be
performed while the cluster is operative, while others require planned
downtime. Make sure you plan all the necessary actions carefully. This will
spare you a lot of trouble.
8.3.1 Nodes
When maintaining or upgrading a node, cluster services must usually be
stopped on the node. This means down time for the applications usually
running on this node, at least during the takeover to other nodes.
Consider the following points when replacing the whole or components of a
node:
• Make sure you have at least the same amount of RAM in the replacement
system.
• If your applications have been optimized for a particular processor or
architecture, ensure that the new node is the same type of system.
Uniprocessor applications may run slower on SMP systems.
• Slot capacity of the new node must be the same or better.
• Check the appropriate documentation for a proper adapter placement in
your new node.
• The license of your application may be dependent on the CPU ID. You
may need to apply for a new license before trying to bring the new node
into service.
8.3.2 Adapters
In order to replace or add an adapter, the node must be powered off. This
means down time for the applications usually running on this node, at least
during the takeover to other nodes.
Consider the following points when replacing or adding adapters in a node:
• Make sure that the adapter is your problem and not faulty cabling. Bad
cables are much more common than defective adapters. Most network
and SSA cables can be changed online. Do some testing, for example,
exchange the cables or try to connect to another port in your hub to see if
the hub is your problem.
160
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• The new adapter must be of the same type or a compatible type as the
replaced adapter.
• When replacing or adding an SCSI adapter, remove the resistors for
shared buses. Furthermore, set the SCSI ID of the adapter to a value
different than 7.
8.3.3 Disks
Disk failures are handled differently according to the capabilities of the disk
type and the HACMP version you are running. Whether your data is still
available after a disk crash, and whether you will need down time to
exchange it, will depend on the following questions:
• Is all the data on the failed disk mirrored to another disk, or is the failed
disk part of a RAID array?
• Will the volume group stay online (Quorum)?
• Is the type of disk you are using hot-swappable?
8.3.3.1 SSA/SCSI Disk Replacement (RAID)
RAID arrays are typically designed for concurrent maintenance. No command
line intervention should be necessary to replace a failed disk in a RAID array.
Do the following steps in order to replace a disk that is a member of a RAID
array:
1. Remove the disk logically from the RAID array (for example with the
appropriate SMIT menu). Removing a disk from a RAID array is known as
reducing the RAID array. No more than one disk can be removed from an
array at one time.
2. Remove the failed disk and plug in the substitute disk.
3. Add the replacement disk logically to the RAID array. All information from
the original disk will be regenerated on the substitute disk. Once data
regeneration has completed on the new disk, the array will return to its
normal optimal mode of operation.
8.3.3.2 Disk Replacement (Non-RAID) before HACMP version 4.3
If LVM mirroring is used, some careful manual steps must be followed to
replace a failed SCSI or SSA disk:
1. Identify which disk has failed, using errpt, lspv, lsvg, diags.
2. Remove all LV copies from the failed disk (rmlvcopy).
3. Remove the disk from the VG (reducevg).
Cluster Management and Administration 161
Download from Www.Somanuals.com. All Manuals Search And Download.
4. Logically remove the disk from the system (rmdev -l hdiskX -d; rmdev -l
pdiskY -d if a SSA disk) on all nodes.
5. Physically remove the failed disk and replace it with a new disk.
6. Add the disk to the ODM (mkdevor cfgmgr)on all nodes.
7. Add the disk to the shared volume group (extendvg).
8. Increase the number of LV copies to span across the new disk (mklvcopy).
9. Synchronize the volume group (syncvg)
Note
Steps 10 and 11 are only necessary in HACMP versions prior to 4.2. With
HACMP 4.2 and later Lazy Update will export/import the volume group on
the backup node in case of a takeover. However, it is necessary to update
the PVID of the replaced disk on the backup nodes manually.
10.Stop all the application(s) using the shared volume group, varyoff the
shared volume group and export/import it on the backup node(s).
Furthermore set the characteristics of the shared volume group
(autovaryon and quorum) on the backup node(s), then vary it off again.
11.Varyon the shared volume group on it’s “normal” node and start the
application(s).
8.3.3.3 Disk Replacement (Non-RAID) with HACMP version 4.3
With the HACMP 4.3 enhancements to the C-SPOC LVM utilities, the disk
replacement does not cause system down time, as long as the failed disk was
part of a RAID array, or if all the LVs on it are mirrored to other disks, and the
failed disk is hot-swappable.
1. Identify which disk has failed using errpt, lspv, lsvg, diag.
2. Remove all LV copies from the failed disk (smit cl_lvsc).
3. Remove the disk from the VG (smit cl_vgsc).
4. Logically remove the disk from the system (rmdev -l hdiskX -d, rmdev -l
pdiskY -dif SSA disk)
5. Physically remove the failed disk and replace it with a new disk.
6. Add the new disk to the ODM (mkdevor cfgmgr).
7. Add the new disk to the sharedvg (smit cl_vgsc).
8. Increase the number of LV copies to span across the new disk (smit
cl_lvsc).
9. Sync the volume group (smit cl_syncvg).
162
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
8.4 Changing Shared LVM Components
Changes to VG constructs are probably the most frequent kind of changes to
be performed in a cluster. As a system administrator of an HACMP for AIX
cluster, you may be called upon to perform any of the following LVM-related
tasks:
• Creating a new shared volume group
• Extending, reducing, changing, or removing an existing volume group
• Importing, mirroring, unmirroring, or synchronizing mirrors of a volume
group
• Creating a new shared logical volume
• Extending, reducing, changing, copying, or removing an existing logical
volume (or a copy)
• Creating a new shared file system
• Extending, changing, or removing an existing file system
The varyon of a shared volume group will only succeed if the information
stored in the VGDA on the disks of the shared volume group and the
information stored in the ODM are equal. After changes in the volume group
(e. g. increasing the size of a file system), the information about the volume
group in ODM and in the VGDA on the disks are still equal, but it will be
different from the information in the ODM of a node that did not have the
volume group varied on at the time of the change. In order to keep a takeover
from failing, the volume group information must be synchronized. There are
four distinct ways to keep all the volume group ODMs synchronized:
• Manual Update
• Lazy Update
• C-SPOC
• TaskGuide
Chapters 4 and 5 of the HACMP for AIX, Version 4.3: Administration Guide,
SC23-4279, describe in detail how to change shared LVM components.
8.4.1 Manual Update
Sometimes, manual updates of shared LVM components are inevitable
because you cannot do some of the tasks mentioned above with any of the
tools. For example, neither with C-SPOC nor with TaskGuide or Lazy Update
is it possible to remove a VG on all of the cluster nodes.
Cluster Management and Administration 163
Download from Www.Somanuals.com. All Manuals Search And Download.
When changing shared LVM components manually, you will usually need to
run through the following procedure:
1. Stop HACMP on the node owning the shared volume group (sometimes a
stop of the applications using the shared volume group may be sufficient).
2. Make the necessary changes to the shared LVM components.
3. Unmount all the file systems of the shared volume group.
4. Varyoff the shared volume group.
5. Export the old volume group definitions on the next node.
6. Import the volume group from one of its disks on the next node. Make sure
you use the same VG major number.
7. Change the volume group to not auto-varyon at system boot time.
8. Mount all the file systems of the shared volume group.
9. Test the file systems.
10.Unmount the file systems of the shared volume group.
11.Varyoff the shared volume group.
12.Repeat steps 6 through 11 for all the other nodes with an old ODM of the
shared volume group.
13.Start HACMP again on the node usually owning the shared volume group.
8.4.2 Lazy Update
For LVM components under the control of HACMP for AIX, you do not have to
explicitly export and import to bring the other cluster nodes up-to-date.
Instead, HACMP for AIX can perform the export and import when it activates
the volume group during a failover. In a cluster, HACMP controls when
volume groups are activated. HACMP for AIX implements a function, called
Lazy Update, by keeping a copy of the timestamp from the volume group’s
VGDA. AIX updates this timestamp whenever the LVM component is
modified. When another cluster node attempts to vary on the volume group,
HACMP for AIX compares its copy of the timestamp (kept in the
/usr/sbin/cluster/etc/vg file) with the timestamp in the VGDA on the disk. If the
values are different, the HACMP for AIX software exports and re-imports the
volume group before activating it. If the timestamps are the same, HACMP for
AIX activates the volume group without exporting and re-importing.
The time needed for takeover expands by a few minutes if a Lazy Update
occurs. A Lazy Update is always performed the first time a takeover occurs in
order to create the timestamp file on the takeover node.
164
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Lazy Update has some limitations, which you need to consider when you rely
on Lazy Update in general:
• If the first disk in a sharedvg has been replaced, the importvgcommand
will fail as Lazy Update expects to be able to match the hdisk number for
the first disk to a valid PVID in the ODM.
• Multi-LUN support on the SCSI RAID cabinets can be very confusing to
Lazy Update as each LUN appears as a new hdisk known to only one
node in the cluster (remember that Lazy Update works on LVM
constructs).
8.4.3 C-SPOC
The Cluster Single Point of Control (C-SPOC) utility lets system
administrators perform administrative tasks on all cluster nodes from any
node in the cluster. These tasks are based on commonly performed AIX
system administration commands that let you:
178).
• Maintain shared Logical Volume Manager (LVM) components.
Without C-SPOC functionality, the system administrator must spend time
executing administrative tasks individually on each cluster node. Using the
C-SPOC utility, a command executed on one node is also executed on other
cluster nodes. Thus C-SPOC minimizes administrative overhead and reduces
the possibility of inconsistent node states. For example, to add a user, you
usually must perform this task on each cluster node. Using C-SPOC,
however, you issue a C-SPOC command once on a single node, and the user
is added to all specified cluster nodes.
C-SPOC also makes managing logical volume components and controlling
cluster services more efficient. You can use the C-SPOC utility to start or stop
cluster services on nodes from a single node.
C-SPOC provides this functionality through its own set of cluster
administration commands, accessible through SMIT menus and screens. To
use C-SPOC, select the Cluster System Management option from the
HACMP for AIX menu. See the HACMP for AIX, Version 4.3: Administration
Guide, SC23-4279 for detailed information on using C-SPOC SMIT options.
With C-SPOC, you can perform the following tasks:
Cluster Management and Administration 165
Download from Www.Somanuals.com. All Manuals Search And Download.
• Shared volume groups
• List all volume groups in the cluster.
• Import a volume group (with HACMP 4.3 only).
• Extend a volume group (with HACMP 4.3 only).
• Reduce a volume group (with HACMP 4.3 only).
• Mirror a volume group (with HACMP 4.3 only).
• Unmirror a volume group (with HACMP 4.3 only).
• Synchronize volume group mirrors (with HACMP 4.3 only).
• Shared logical volumes
• List all logical volumes by volume group.
• Add a logical volume to a volume group (with HACMP 4.3 only).
• Make a copy of a logical volume.
• Remove a copy of a logical volume.
• Show the characteristics of a logical volume.
• Set the characteristics of a logical volume (name, size); this is only
possible in non-concurrent mode and with HACMP 4.3.
• Remove a logical volume.
• Shared file systems (only applicable for non-concurrent VGs)
• List all shared file systems.
• Change/View the characteristics of a shared file system.
• Remove a shared file system.
C-SPOC has the following limitations:
• C-SPOC does not offer the option for creating volume groups. Use the
TaskGuide or standard AIX commands.
• The new Volume Group must be imported manually to other nodes in the
resource group (TaskGuide does it automatically).
• The Volume Group must be defined in a resource group, and cluster
resources must be synchronized, prior to using C-SPOC to manage it.
• C-SPOC does not offer an option for creating file systems. Use standard
AIX commands or SMIT menus to create file systems, and use C-SPOC to
update the VG information on the other nodes.
• C-SPOC cannot be used for concurrent shared LVM components prior to
HACMP 4.3 for AIX.
166
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
To use the SMIT shortcuts to C-SPOC, type smit cl_lvm or smit cl_conlvm for
concurrent volume groups. Concurrent volume groups must be varied on in
concurrent mode to perform tasks.
8.4.4 TaskGuide
The TaskGuide is a graphical interface that simplifies the task of creating a
shared volume group within an HACMP cluster configuration. The TaskGuide
presents a series of panels that guide the user through the steps of specifying
initial and sharing nodes, disks, concurrent or non-concurrent access, volume
group name and physical partition size, and cluster settings. The TaskGuide
can reduce errors, as it does not allow a user to proceed with steps that
conflict with the cluster’s configuration. Online help panels give additional
information to aid in each step.
8.4.4.1 TaskGuide Requirements
TaskGuide is only available since HACMP for AIX version 4.3. Before you
start the TaskGuide, make sure that:
• You have a configured HACMP cluster in place
• You are on a graphics capable terminal
8.4.4.2 Starting the TaskGuide
You can start the TaskGuide from the command line by typing:
/usr/sbin/cluster/tguides/bin/cl_ccvg, or you can use the SMIT interface as
follows:
1. Type smit hacmp
2. From the SMIT main menu, choose Cluster System Management ->
Cluster Logical Volume Manager ->Taskguide for Creating a Shared
Volume Group. After a pause, the TaskGuide Welcome panel appears.
3. Proceed through the panels to create or share a volume group.
8.5 Changing Cluster Resources
In HACMP for AIX, you define each resource as part of a resource group.
This allows you to combine related resources into a single logical entity for
easier configuration and management. You then configure each resource
group to have a particular kind of relationship with a set of nodes. Depending
on this relationship, resources can be defined as one of three types:
cascading, rotating, or concurrent access. You also assign a priority to each
participating node in a cascading resource group chain.
Cluster Management and Administration 167
Download from Www.Somanuals.com. All Manuals Search And Download.
To change the nodes associated with a given resource group, or to change
the priorities assigned to the nodes in a resource group chain, you must
redefine the resource group. You must also redefine the resource group if you
add or change a resource assigned to the group. This section describes how
to add, change, and delete a resource group.
8.5.1 Add/Change/Remove Cluster Resources
You can add, change and remove a resource group in an active cluster. You
do not need to stop and then restart cluster services for the resource group to
become part of the current cluster configuration.
Use the following SMIT shortcuts:
To add a resource group, use smit cm_add_grp.
To remove a resource group, use smit cm_add_res.
To change a resource group, use smit cm_add_res.
Whenever you modify the configuration of cluster resources in the ODM
on one node, you must synchronize the change across all cluster nodes.
8.5.2 Synchronize Cluster Resources
You perform a synchronization by choosing the Synchronize Cluster
Resources option from the Cluster Resources SMIT screen.
Note
In HACMP for AIX, the event customization information stored in the ODM
is synchronized across all cluster nodes when the cluster resources are
synchronized. Thus, pre, post, notify, and recovery event script names
must be the same on all nodes, although the actual processing done by
these scripts can be different.
The processing performed in synchronization varies depending on whether
the Cluster Manager is active on the local node:
• Ιf the cluster manager is not active on the local node when you select this
option, the ODM data in the DCD (Default Configuration Directory–for
more information, see Chapter 3 in the HACMP for AIX, Version 4.3:
Concepts and Facilities, SC23-4276) on the local node is copied to the
ODMs stored in the DCDs on all cluster nodes.
168
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
• If the Cluster Manager is active on the local node, synchronization triggers
a cluster-wide, dynamic reconfiguration event. In dynamic reconfiguration,
the configuration data stored in the DCD is updated on each cluster node,
and, in addition, the new ODM data replaces the ODM data stored in the
ACD (Active Configuration Directory) on each cluster node. The cluster
daemons are refreshed and the new configuration becomes the active
configuration. In the HACMP for AIX log file, reconfig_resource_release,
reconfig_resource_acquire, and reconfig_resource_complete events mark
the progress of the dynamic reconfiguration.
• If the Cluster Manager is active on some cluster nodes but not on the local
node, the synchronization is aborted.
8.5.3 DARE Resource Migration Utility
The HACMP for AIX software provides a Dynamic Reconfiguration (DARE)
Resource Migration utility that allows for improved cluster management by
allowing a system administrator to alter the placement of resource groups
(along with their resources—IP addresses, applications, and disks) to specific
cluster nodes using the cldarecommand. The command lets you move the
ownership of a series of resource groups to a specific node in that resource
group’s node list, as long as the requested arrangement is not incompatible
with the current resource group configuration. It also lets you disable
resource groups, preventing them from being acquired during a failover or
reintegration.
Dynamic resource group movement essentially lets a system administrator
better use hardware resources within the cluster, forcing resource traffic onto
one or more high-powered or better-connected nodes without having to shut
down HACMP on the node from which the resource group is moved. Dynamic
resource group movement also lets you perform selective maintenance
without rebooting the cluster or disturbing operational nodes.
Using the DARE Resource Migration utility does not affect other resource
groups that might currently be owned by that node. The node that currently
owns the resource group will release it as it would during a “graceful
shutdown with takeover”, and the node to which the resource group is being
moved will acquire the resource group as it would during a node failover.
The following section covers the types and location keywords used in DARE
resource migrations, and also how to use the cldare command and the -M
flag to perform the migration.
Cluster Management and Administration 169
Download from Www.Somanuals.com. All Manuals Search And Download.
8.5.3.1 Resource Migration Types
Before performing a resource migration, decide if you will declare the
migration stickyor non-sticky.
Sticky Resource Migration
A sticky migration permanently attaches a resource group to a specified
node. The resource group attempts to remain on the specified node during a
node failover or reintegration.
Since stickiness is a behavioral property of a resource group, assigning a
node as a sticky location makes the specified resource group a sticky
resource. Older sticky locations are superseded only by new sticky migration
requests for the same resource group, or they are removed entirely during
non-sticky migration requests for the same resource group. If it is not
possible to place a resource group on its sticky location (because that node is
down), the normal resource policy is invoked, allowing the resource to
migrate according to the takeover priority specified in the resource group’s
node list.
For both cascading and rotating resource groups, a normal resource policy
means that other cluster nodes in the group’s node list are consulted at the
time the sticky location fails to find the highest-priority node active. After
finding the active node, cascading resource groups will continually migrate to
the highest-priority node in the group’s node list (ultimately residing at the
sticky location). Rotating resource groups stay put until the sticky location
returns to the cluster.
You can attach the optional keyword stickyto any migration you perform,
regardless of the resource group configuration (rotating or cascading).
However, with very few exceptions, you always use the sticky location for
cascading configurations, and do not use it for rotating configurations.
Non-Sticky Resource Migration
Resource groups on nodes not designated sticky are by default transient,
non-sticky resources. These resources are temporarily placed on the
specified node with the highest priority in the node list until the next failover or
reintegration occurs. Non-sticky resources are best suited for use with
rotating resource group configurations because of this transient behavior.
Because the normal behavior of cascading resources is to bound back to the
highest available node in their node list, non-sticky migrations are usually not
the best choice. The one instance in which a non-sticky migration of a
cascading resource might make sense is if this resource has the
170
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
INACTIVE_TAKEOVER flag set to false and has not yet started because its
primary node is down.
In general, however, only rotating resource groups should be migrated in a
non-sticky manner. Such migrations are one-time events and occur similar to
normal rotating resource group flavors. After migration, the resource group
immediately resumes a normal rotating resource group failover policy, but
from the new location.
Note
The cldarecommand attempts to perform all requested migrations
simultaneously. If, for some reason, the command cannot simultaneously
cause all specified resources to be released and cannot simultaneously
reacquire them at the new locations, it fails, and no migrations occur.
8.5.3.2 Locations
You can specify the location for a resource group by entering a node name or
a keyword.
Node Name
In most cases, you enter a node name in the location field to specify which
node will contain sticky or non-sticky resource groups. Node names can be
arbitrary and apply to both rotating and cascading resource group
configurations.
The DARE Resource Migration utility also provides the following special
keywords you can use in the location field to determine the placement of
migrated resource groups: defaultand stop. The defaultand stoplocations
are special locations that determine resource group behavior and whether the
resources can be reacquired.
Default Location
If you use the default keyword as the location specifier, the DARE Resource
Migration utility removes all previous stickiness for the resource group and
returns the resource group to its default failover behavior where node
priorities apply (for either cascading or rotating resources). The use of a
default destination for a cascading resource group returns it to its normal
behavior (the resource group will migrate to the highest priority node currently
up). Using a default destination for a rotating resource group releases the
group from wherever it resides and lets the highest priority node with a boot
address reacquire the resource.
Cluster Management and Administration 171
Download from Www.Somanuals.com. All Manuals Search And Download.
If you do not include a location specifier in the location field, the DARE
Resource Migration utility performs a default migration, again making the
resources available for reacquisition.
Note
A default migration can be used to start a cascading resource group that
has INACTIVE_TAKEOVER set to false and that has not yet started
because its primary node is down.
Stop Location
The second special location keyword, stop, causes a resource group to be
made inactive, preventing it from being reacquired, though it remains in the
resource configuration. Its resources remain unavailable for reacquisition
even after a failover or reintegration.
8.5.3.3 Using the cldare Command to Migrate Resources
The cldarecommand can be used to perform dynamic resource group
migrations to other cluster nodes in conjunction with other cldare resource
functionality. It lets you specify multiple resource groups and nodes on the
command line, as long as the final resource group configuration is consistent.
After some error checking, the resources are released and reacquired by the
specified cluster nodes. Resource migration first releases all specified
resources (wherever they reside in the cluster); then it reacquires these
resources on the newly specified nodes.
You can also use this command to swap resources on nodes in the resource
group’s node list, but you cannot mix keywords—default, stop,and
node—when using the cldarecommand.
To migrate resource groups (and their resources) using the cldarecommand,
enter the following command:
cldare -M <resgroup name>:[location|[default|stop]][:sticky] ...
where -Mspecifies migration, and where resource group names must be valid
names of resource groups in the cluster. You can specify a node name (or
special location) or the keyword stopor defaultafter the first colon. The node
name must represent a cluster node that is up and in the resource group’s
node list. You can specify a migration type after the second colon. Repeat this
syntax on the command line for each resource group you want to migrate. Do
not include spaces between arguments.
172
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Note that you cannot add nodes to the resource group list with the DARE
Resource Migration utility. This task is performed through SMIT.
Stopping Resource Groups
If the location field of a migration contains the keyword stopinstead of an
actual nodename, the DARE Resource Migration utility attempts to stop the
resource group, which includes taking down any service label, unmounting
file systems, and so on. You should typically supplement the keyword stop
with the migration type stickyto indicate that the resource stays down, even
if you reboot the cluster.
As with sticky locations, sticky stop requests are superseded by new sticky
migration requests for the same resource group, or they are removed by
default,non-sticky migration requests for the same resource group. Thus, a
stopped resource will be restarted at the time of the next migration request.
Note
Be careful when using a non-sticky stop request, since the resource group
will likely be restarted at the next major cluster event. As a result, all
non-sticky requests produce warning messages. A non-sticky stop could
be used to halt a cascading resource group that has
INACTIVE_TAKEOVER set to false during periods in which its primary
node is down.
8.5.3.4 Using the clfindres Command
To help you locate resources placed on a specific node, the DARE Resource
Migration utility includes a command, clfindres, that makes a best-guess
estimate (within the domain of current HACMP configuration policies) of the
state and location of specified resource groups. It also indicates whether a
resource group has a sticky location, and it identifies that location.
See Appendix A of the HACMP for AIX, Version 4.3: Administration Guide,
SC23-4279, for the syntax and typical output of the clfindrescommand.
8.5.3.5 Removing Sticky Markers When the Cluster is Down
Sticky location markers are stored in the HACMPresource class in the
HACMP ODM and are a persistent cluster attribute. While the cluster is up,
you can only remove these locations by performing a subsequent non-sticky
migration on the same resource group, using the defaultspecial location
keyword or specifying no location.
Cluster Management and Administration 173
Download from Www.Somanuals.com. All Manuals Search And Download.
Be aware that persistent sticky location markers are saved and restored in
cluster snapshots. You can use the clfindrescommand to find out if sticky
markers are present in a resource group.
If you want to remove sticky location markers while the cluster is down, the
defaultkeyword is not a valid method, since it implies activating the resource.
Instead, when the cluster is down, you use a transient stoprequest, as in this
example:
cldare -v -M <resgroup name>:stop
(The optional -v flag indicates that verification is skipped.)
8.6 Applying Software Maintenance to an HACMP Cluster
You can install software maintenance, called Program Temporary Fixes
(PTFs), to your HACMP cluster while running HACMP for AIX cluster services
on cluster nodes; however, you must stop cluster services on the node on
which you are applying a PTF. As with everything else in a cluster, applying
software fixes should be done in a controlled fashion.
With the method described below, you might even be able to keep your
mission-critical application up and running during the update process,
provided that the takeover node is designed to carry its own load and the
takeover load as well.
The normal method of applying AIX fixes is to do the following:
1. Use the smit clstopfastpath to stop cluster services on the node on which
the PTF is to be applied. If you would like the resources provided by this
node to remain available to users, stop cluster with takeover so that the
takeover node will continue to provide these resources to users.
2. Apply the software maintenance to this node using the procedure
described in the documentation distributed with the PTF.
3. Run the /usr/sbin/cluster/diag/clverifyutility to ensure that no errors
exist after installing the PTF. Test the fix as thoroughly as possible.
4. Reboot the node to reload any HACMP for AIX kernel extensions that may
have changed as a result of the PTF being applied.
If an update to the cluster.base.client.lib file set has been applied and you
are using Cluster Lock Manager or Clinfo API functions, you may need to
relink your applications.
174
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
5. Restart the HACMP for AIX software on the node using the smit clstart
fastpath and verify that the node successfully joined the cluster.
6. Repeat Steps 1 through 5 on the remaining cluster nodes.
Figure 15 below shows the procedure:
Fallover of
System A
System A
rejoins the
cluster
System A
System B
System A
System B
System A
System B
Apply PTFs
to System A
Test
System A
System A
System B
System A's
Resource
System B's
Resource
Figure 15. Applying a PTF to a Cluster Node
Along with the normal rules for applying updates, the following general points
should be observed for HACMP clusters:
• Cluster nodes should be kept at the same AIX maintenance levels
wherever possible. This will, of course, not be true while the update is
being applied, but should be true at all other times.
Cluster Management and Administration 175
Download from Www.Somanuals.com. All Manuals Search And Download.
• Cluster nodes should be running the same HACMP maintenance levels.
There might be incompatibilities between various maintenance levels of
HACMP, so you must ensure that consistent levels are maintained across
all cluster nodes. The cluster must be taken down to update the
maintenance levels.
8.7 Backup Strategies
HACMP software masks hardware failures in clustered RISC System/6000
environments by quickly switching over to backup machines or other
redundant components. However, installing HACMP is not a substitute for
having a formal backup and recovery procedure.
In general, a backup of user and system data is kept in case data is
accidentally removed or in case of a disk failure. A formal backup process is
really an insurance policy. You invest in the technology and time to back up
systems so that, in the event of a problem, you can quickly rebuild the
system.
Since system and application backups are preferably done during periods of
no usage (for instance, in the middle of the night), many installations
implement an automated backup procedure using the AIX cron facility. While
this is a very good procedure, the HACMP cluster environment presents
some special challenges. The problem is, you never know which machine has
your application data online, so you need to ensure that exactly the node that
has a resource online will initiate the backup of data.
It isn’t actually important which of the several backup commands you are
using, what is important is the strategy. For the features and/or restrictions of
backup commands like tar, cpio, dd or backup, refer to the AIX Commands
Reference Version 4.3, SBOF-1877.
8.7.1 Split-Mirror Backups
No file system can be safely backed up while update activity is occurring. If
you are going to have any assurance as to which updates are on the backup
and which updates are not, you need to be able to demark exactly where the
backup was made. Therefore, it may be difficult to do a good backup on
systems that have applications or data that must be online continuously or
offline for only a very short time. In some installations, the time required to do
a full backup to an archival device, or even to another, might be longer than
the availability requirements of the application will allow it to be offline. The
mirroring capability of the AIX Logical Volume Manager (LVM) can be used to
address this issue.
176
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
8.7.1.1 How to do a split-mirror backup
This same procedure can be used with just one mirrored copy of a logical
volume. If you remove a mirrored copy of a logical volume (and file system),
and then create a new logical volume (and file system) using the allocation
map from that mirrored copy, your new logical volume and file system will
contain the same data as was in the original logical volume.
Now, you can mount this new file system (read-only is recommended), back it
up, and you are really backing up a mirrored copy of the data in the original
file system, as it was when we removed the mirror copy. Since this file
system, created from the mirror copy, is mounted read-only, no inconsistency
in the file system from the point at which you removed the mirror originally is
created during the backup. After that, you can delete the new file system to
release the physical partitions back to the free pool. Finally, you can add and
synchronize a mirror back onto the original file system, and you are back to a
mirrored mode of operation, with fully updated data.
The splitlvcopycommand of AIX does much of the work required to
implement this solution.
We can summarize the steps to do a split-mirror backup of a file system as
follows:
1. Use the lsvg -l VGNAME command to take note of the logical volumet name
that contains the file system you want to back up.
2. Stop any application using the file system and unmount the file system.
3. Use the splitlvcopycommand to break off one mirror of the logical
volume, and create a new logical volume with its contents. For example, if
the existing logical volume is named fslv, the command would be
splitlvcopy -y newlv fslv.
4. It is important to note that there is now one less mirror copy available to
the user in fslv.
5. Remount the file system and restart the application that was using it.
6. You can see that the application and data are offline for only a very short
time.
7. Create a file system on your new logical volume and mount it read-only.
This is to ensure that no update activity will occur in the new logical
volume, and the consistency of the file system is guaranteed.
8. Perform the backup on the new file system by any means desired, such as
backup, tar, cpio,and pax.
Cluster Management and Administration 177
Download from Www.Somanuals.com. All Manuals Search And Download.
9. After the backup is complete and verified, unmount and delete the new file
system and the logical volume you used for it.
10.Use the mklvcopycommand to add back the logical volume copy you
previously split off to the fslv logical volume.
11.Resynchronize the logical volume.
Once the mirror copy has been recreated on the logical volume, the syncvg
command will resynchronize all physical partitions in the new copy, including
any updates that have occurred on the original copy during the backup
process.
It is always a good idea to check a backup for validity.
8.7.2 Using Events to Schedule a Backup
As described above, a crontabentry is often used for scheduling nightly
backups during off-peak hours of the application. Now as you have several
cluster nodes, each of them would need a crontabentry, in order to get its
own data backed up. This crontabentry can determine whether only the
“normal” data is backed up, i.e. the data this cluster node cares about during
“normal” operations, or, in case of another’s node failure and a subsequent
takeover of this node’s resources, backing up both of the cluster nodes’ data.
Whenever one node takes over the reources of another node, the
node_down_remote event has happened. You can use a post-event to the
node_down_remote event to change the crontabentry from backing up only
the local node’s data into backing up both nodes’ data.
Furthermore, if the second node eventually comes up again and takes its
resources back, you will see a node_up_remote event in your logs. Thus, you
can configure a post-event to the node_up_remote event to change the
crontab entry back to the “normal” setting.
If you want to do a split-mirror backup, the crontab entry has to invoke a
script, implementing the steps described above.
A more detailed description of this procedure can be found in the redbook
HACMP/6000 Customization Examples, SG24-4498, Chapter 6.
8.8 User Management
As 2.7, “User ID Planning” on page 48 described, on an HACMP cluster, the
administrator has to take care of user and group IDs throughout the cluster. If
178
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
they don’t match, the user won’t get anything done after a failover happened.
So, the administrator has to keep definitions equal throughout the cluster.
Fortunately, the C-SPOC utility, as of HACMP Version 4.3 and later, does this
for you. When you create a cluster group or user using C-SPOC, it makes
sure that it has the same group id or user id throughout the cluster.
8.8.1 Listing Users On All Cluster Nodes
To obtain information about all user accounts on cluster nodes (or about a
particular user account), you can either use the AIX lsusercommand in rshto
one cluster node after another, or use the C-SPOC cl_lsusercommand, or
the C-SPOC SMIT List all the Users on the Cluster screen. The cl_lsuser
command executes the AIX lsusercommand on each node. To obtain a
listing of all user accounts in the cluster, you must specify the ALLargument.
If you specify a user name that does not exist on one of the cluster nodes, the
cl_lsusercommand outputs a warning message but continues execution of
the command on other cluster nodes.
Note
If you have a Network Information Service (NIS) database installed on any
cluster node, some user information may not appear when you use the
cl_lsusercommand.
8.8.2 Adding User Accounts on all Cluster Nodes
Adding a user to the cluster involves three steps:
1. Add an entry for the new user to the /etc/passwd file and other system
security files.
2. Create a home directory for the new user.
3. Add the user to a group file.
On AIX systems, you use the mkusercommand to perform these tasks. This
command adds entries for the new user to various system security files,
including /etc/passwd and /etc/security/passwd, adds the new user to a
group, and creates a home directory for the new user. Every user account
has a number of attributes associated with it. When you create a user, the
mkusercommand fills in values for these attributes from the system default
/usr/lib/security/mkuser.default file. You can override these default values by
specifying an attribute and a value on the mkusercommand line.
Cluster Management and Administration 179
Download from Www.Somanuals.com. All Manuals Search And Download.
To add a user on one or more nodes in a cluster, you can either use the AIX
mkusercommand in a rshto one clusternode after the other, or use the
C-SPOC cl_mkusercommand or the Add a User to the Cluster SMIT screen.
The cl_mkusercommand calls the AIX mkusercommand to create the user
account on each cluster node you specify. The cl_mkusercommand creates a
home directory for the new account on each cluster node.
8.8.3 Changing Attributes of Users in a Cluster
On AIX systems, you can change any of the attributes associated with an
existing user account by using the chusercommand. Using the chuser
command, you specify the name of the user account you want to change and
then specify the attributes with their new values. If you use the SMIT Change
User Attributes screen, the complete list of user attributes is displayed and
you can supply new values for any attributes. The chusercommand modifies
the user information stored in the /etc/passwd file and the files in the
/etc/security directory.
To change the attributes of a user account on one or more cluster nodes, you
can either use the AIX chusercommand in rshto one cluster node after the
other, or use the C-SPOC cl_chusercommand or the C-SPOC Change User
Attributes SMIT screen. The cl_chusercommand executes the AIX chuser
command on each cluster node.
Note
Do not use the cl_chusercommand if you have an NIS (Network
Information Service) database installed on any node in your cluster.
Both cluster nodes must be active and a user with the specified name must
exist on both the nodes for the change operation to proceed. Optionally, you
can specify that the cl_chusercommand continue processing if the specified
user name exists on any of the cluster nodes. See the cl_chusercommand
man page for more information.
8.8.4 Removing Users from a Cluster
On AIX systems, you remove a user account by using the rmusercommand or
the SMIT Remove a User From the System screen. Using the rmuser
command you specify the name of the user account you want to remove and
specify whether you want the user password and other authentication
information removed from the /etc/security/passwd file.
180
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
To remove a user account from one or more cluster nodes, you can either use
the AIX rmusercommand on one cluster node after the other, or use the
C-SPOC cl_rmusercommand or the C-SPOC Remove a User from the
Cluster SMIT screen. The cl_rmusercommand executes the AIX rmuser
command on all cluster nodes.
Note
The system removes the user account but does not remove the home
directory or any files owned by the user. These files are only accessible to
users with root authority or by the group in which the user was a member.
8.8.5 Managing Group Accounts
In order to manage a number of similar users as a single entity, AIX provides
the administrator with the group concept. Members of one group share the
same permissions, the same attributes and limits, and so on.
Commands for managing group accounts are just like the user managing
commands very much alike to the native AIX commands. The restrictions on
NIS are just the same as for users, and therefore are not explained here in
detail.
For more detailed information, please refer to Chapter 12 of the HACMP for
AIX, Version 4.3: Administration Guide, SC23-4279.
8.8.6 C-SPOC Log
Because these commands are running and executing while distributed
amongst the cluster, it could happen that something doesn’t work exactly like
it should. The C-SPOC utility, therefore, maintains a log on the initiating node.
It can be found under /tmp/cspoc.log.
Note that the initiating node doesn’t have to be the same in all cases, so the
log file might be present on different cluster nodes, and doesn’t contain the
same data.
Cluster Management and Administration 181
Download from Www.Somanuals.com. All Manuals Search And Download.
182
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 9. Special RS/6000 SP Topics
This chapter will introduce you to some special topics that only apply if you
are running HACMP on the SP system.
9.1 High Availability Control Workstation (HACWS)
If you are thinking about what could happen to your SP whenever the Control
Workstation might fail, you will probably think about installing HACWS for
that. These paragraphs will not explain HACWS in full detail, but will
concentrate on the most important issues for installation and configuration.
For more details, refer to Chapter 3, “Installing and Configuring the High
Availability Workstation”, in the IBM Parallel System Support Programs for
AIX Installation and Migration Guide, GA22-7347, or to Chapter 4,”Planning
for a High Availability Workstation”, in the IBM RS/6000 SP Planning Volume
2, Control Workstation and Software Environment, GA22-7281.
Some services of the control workstation (or cws for short) are vital, so a
failure would impact your ability to manage an SP system. Also, the failure of
the control workstation could cause the switch network to fail. HACWS covers
the following cases with a fully functional environment:
• Continues running your SP system after a cws failure
• Shuts down the cws for deferred hardware and software maintenance
without having a system outage
• Maintains the SP system function and reliability when the cws fails
• Fails over the cws to a backup
9.1.1 Hardware Requirements
To build a cluster consisting of two control workstations, you have to think
about shared resources. The spdata file system holding the SDR data and
other vital data has to be accessible from both control workstations, so, it has
to be put onto a shared disk.
The cws connects to the frames of an RS/6000 SP with RS232 lines as its
supervisor network. If the RS/6000 SP consists of multiple frames, you will
probably have an 8-port adapter installed in the Control Workstation in order
to provide the needed number of ttys.
To connect a backup cws to the frames, you need exactly the same tty port
configuration as on the primary cws, that is, when frame 3 connects to tty3 on
the primary cws, it has to connect to tty3 on the backup cws as well. Also, you
© Copyright IBM Corp. 1999
183
Download from Www.Somanuals.com. All Manuals Search And Download.
need to have the frame supervisors support dual tty lines in order to get both
control workstations connected at the same time. Contact your IBM
Both the tty network and the RS/6000 SP internal ethernet are extended to
the backup cws. In contrast to standard HACMP, you don’t need to have a
second ethernet adapter on the backup cws. In case you have only one, the
HACWS software will work with ip aliasing addresses on one adapter.
Primary CWS
SP Ethernet
e
l
ab
C
ASD
D
Ac
e
v
t
i
v
Mirror
e
Acti
t
t
HACMP
Serial
Link
y
C
a
HACMP
Heartbeat
b
l
Frame
e
Supervisor
Y-cable
Inacti
v
e
D
le
Cab
ASD
tty
e
tiv
Cab
le
c
a
n
I
SP Ethernet
Backup CWS
Figure 16. A Simple HACWS Environment
9.1.2 Software Requirements
Both of the control workstations must have the same software installed, that
is, they must be on the same AIX level, use the same PSSP software level
and have to have HACMP on the same level as well. For example, if you want
to use HACMP 4.3 for AIX, you have to use PSSP 3.1 and therefore AIX
Version 4.3.2 on the primary cws and on the backup cws.
9.1.3 Configuring the Backup CWS
The primary cws is configured exactly as usual, as far as the AIX and PSSP
software is concerned, as if there were no HACWS at all.
184
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
The backup cws has to be installed with the same level of AIX and PSSP.
Depending on the kerberos configuration of the primary cws, the backup cws
has to be configured either as a secondary authentication server for the
authentication realm of your RS/6000 SP when the primary cws is an
authentication server itself, or as an authentication client when the primary
cws is an authentication client of some other server. To do so will enable a
correct kerberos environment on the backup cws; so, remote commands will
succeed through kerberos authentication as on the primary cws.
After the initial AIX and PSSP setup is done, the HACWS software has to be
installed.
9.1.4 Install High Availability Software
On both control workstations, the HACMP software has to be installed now,
according to the instructions in the HACMP for AIX, Version 4.3: Installation
Guide, SC23-4278. Verification, as described in Chapter 10 of the HACMP for
AIX, Version 4.3: Installation Guide, SC23-4278, should be performed. For
HACWS control workstations, the ssp.hacws fileset has to be installed as
well.
9.1.5 HACWS Configuration
Since the cws might have some daemons active that could interfere with the
definition and configuration of networks, you have to stop them, in order to
get the configuration done, with the command:
/usr/sbin/hacws/spcw_apps -d
This will stop the subsystems spmgr splogd hardmon sysctld supfilesrv
sp_configdif they have been active with the corresponding SRC command.
Now configure the serial network. You can either use target mode SCSI,
target mode SSA or the raw RS-232 serial line, or any combination.
Both machines, primary and backup, need to be configured to boot up on
their boot address in order to not confuse a working cws at boot time of the
backup cws.
If not previously done, you have to migrate the /spdata file system to an
external volume group, to make it accessible from both sides.
After the /spdata file system is set up so that a varyonvgof its vg will work on
either cws, you have to complete the Administration Tasks, like on an
Special RS/6000 SP Topics 185
Download from Www.Somanuals.com. All Manuals Search And Download.
ordinary HACMP cluster, as it is described in Chapter 7 of the HACMP for
AIX, Version 4.3: Installation Guide, SC23-4278.
Now the cluster environment has to be configured. Define a cluster ID and
name for your HACWS cluster and define the two nodes to HACMP.
Adapters have to be added to your cluster definition as described before. You
will have to add a boot adapter and a service adapter for both primary and
backup cws. Now that the cluster topology is defined to HACMP, you can
configure Network Modules as in the HACMP for AIX Installation Guide,
SC23-4278, and synchronize the definitions across the two cluster nodes.
With the ssp.hacws filesets comes a predefined start- and stop-script, which
has to be defined to HACMP as part of the application server definition, which
in turn has to be included in a resource group.
Recommended settings for this resource group are:
Resource Group Name
Node Relationship
[hacws_group1]
[rotating]
Participating Node Names [“nodename of primary cws” “nodename of
backup cws”]
Service IP label
File System
at least the hostname of the primary cws
the name of the file system, most probably
/spdata
Volume Groups
the name of the shared volume group containing
/spdata
Application Servers
the name you gave the application server before
9.1.6 Setup and Test HACWS
Both the primary and backup cws have to be addressable by their hostname,
in order to finish the configuration and check that everything is in order. So,
check if the primary cws can address the backup cws by its hostname and
vice versa. If not, use the ifconfigcommand to temporarily set the interface
to the hostname on each cws. Do NOT use smit chinetfor this, since this
would be a permanent change.
Run the command:
/usr/sbin/hacws/install_hacws -p primary_hostname -b backup_hostname -s
on the primary cws to set up HACWS with the 2 node names.
186
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
After that, identify the HACWS event scripts to HACMP by executing the
/usr/sbin/hacws/spcw_addevents
command, and verify the configuration with the
/usr/sbin/hacws/hacws_verify
command. You should also check the cabling from the backup cws with the
/usr/sbin/hacws/spcw_verify_cabling
command. Then reboot the primary and the backup cws, one after the other,
and start cluster services on the primary cws with smit clstart. After cluster
services is up and running, check that control workstation services, such as
SDRGetObjects, are working as expected. If everything is fine, start up cluster
services on the backup cws as well. Check for the completion of the cluster
services startup with the following command:
grep "SPCW_APPS COMPLETE" /tmp/hacmp.out
Now you can cause a failover by stopping cluster services on the primary cws
and see whether cws services are still available afterwards.
9.2 Kerberos Security
To understand security, we have to clarify some definitions first.
Identification is the process by which an entity tells another who it is.
Authentication is the process by which the other entity verifies this identity.
Authorization is the process performed by an entity to check if an agent,
whose identity has previously been authenticated, has or
does not have the necessary privileges to carry out some
action.
Additionally, if information is transferred over an insecure network, as any
TCP/IP network basically is, there is always a chance that someone is
listening, so some sort of encryption is required.
These issues are solved with kerberos.
Special RS/6000 SP Topics 187
Download from Www.Somanuals.com. All Manuals Search And Download.
Kerberos
Also spelled Cerberus - The watchdog of Hades, whose duty was to guard
the entrance (against whom or what does not clearly appear); it is known to
have had three heads.
- Ambrose Bierce, The Enlarged Devil’s Dictionary
The following is simply a shortened description on how kerberos works. For
more details, the redbook Inside the RS/6000 SP, SG24-5145, covers the
subject in much more detail.
When dealing with authentication and Kerberos, three entities are involved:
the client, who is requesting service from a server; the second entity, and the
Key Distribution Center or Kerberos server, which is a machine that manages
the database, where all the authentication data is kept and maintained.
Kerberos is a third-party system used to authenticate users or services that
are known to Kerberos as principals. The very first action to take regarding
Kerberos and principals is to register the latter to the former. When this is
done, Kerberos asks for a principal’s password, which is converted to a
principal (user or service) 56-bit key using the DES (Data Encryption
Standard) algorithm. This key is stored in the Kerberos server database.
When a client needs the services of a server, the client must prove its identity
to the server so that the server knows to whom it is talking.
Tickets are the means the Kerberos server gives to clients to authenticate
themselves to the service providers and get work done on their behalf on the
services servers. Tickets have a finite life, known as the ticket life span.
In Kerberos terms, to make a Kerberos authenticated service provider work
on behalf of a client is a three-step process:
• Get a ticket-granting ticket.
• Get a service ticket.
• Get the work done on the service provider.
The main role of the ticket-granting ticket service is to avoid unnecessary
password traffic over the network; so, the user should issue his password
only once per session. What this ticket-granting ticket service does is to give
the client systems a ticket that has a certain time span, whose purpose is to
188
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
allow the clients to get service tickets to be used with other servers without
the need to give them the password every time they request services.
So, given a user has a ticket-granting ticket, if a user requests a kerberized
service, he has to get a service ticket for it. In order to get one, the kerberized
command sends an encrypted message, containing the requested service
name, the machine’s name, and a time-stamp to the Kerberos server. The
Kerberos server decrypts the message, checks whether everything is in
order, and if so, sends back a service ticket encrypted with the service’s
private key, so that only the requested service can decrypt it. The client sends
his request along with the just received ticket to the service provider, who in
turn decrypts and checks authorization, and then, if it is in order, provides the
requested service to the client.
9.2.1 Configuring Kerberos Security with HACMP Version 4.3
With HACMP Version 4.3 there is a handy script to do the kerberos setup for
you, called cl_setup_kerberos. It sets up all the IP labels defined to the
HACMP cluster together with the needed kerberos principals, so that remote
kerberized commands will work.
On an SP the setup_authentcommand does the SP-related kerberos setup,
which is based on the IP labels found in the SDR. Since the SDR does not
allow multiple IP labels to be defined on the same interface, whereas HACMP
needs to have multiple IP labels on one interface during IPAT, the kerberos
setup for HACMP has to be redone, every time the setup_authentcommand is
run explicitly or implicitly through the setup_servercommand.
You can either do that manually, or use the cl_setup_kerberostool. To
manually add the kerberos principals, use the kadmincommand. Necessary
principals for kerberized operation in enhanced security mode are the
(remote) rcmd principals and the godm principals. As always, a kerberos
principal consists of a name, godm for example, an IP label, like
hadave1_stby and a realm, so that the principal in its full length would look
like [email protected].
Now after adding all the needed principals to the kerberos database, you
must also add them to the /etc/krb-srvtab file on the nodes. To do that, you
will have to extract them from the database and copy them out to the nodes,
replacing their kerberos file.
Now you can extend root’s .klogin file and /etc/krb.realms file to reflect the
new principals, and copy these files out to the node as well.
Special RS/6000 SP Topics 189
Download from Www.Somanuals.com. All Manuals Search And Download.
After setting the cluster’s security settings to enhanced for all these nodes,
you can verify that it is working as expected, for example, by running clverify,
which goes out to the nodes and checks the consistency of files.
9.3 VSDs - RVSDs
VSDs (Virtual Shared Disks) and RVSDs (Recoverable Virtual Shared Disks)
are SP-specific facilities that you are likely to use in an HACMP environment.
9.3.1 Virtual Shared Disk (VSDs)
Virtual Shared Disk (VSD) allows data in logical volumes on disks physically
connected to one node to be transparently accessed by other nodes.
Importantly, VSD supports only raw logical volumes, not file systems. The
VSD facility is included in the ssp.csd.vsd fileset of PSSP.
IBM developed VSD to enable Oracle’s parallel database on the SP. Oracle’s
database architecture is strongly centralized. Any processing element, or
node, must be able to “see” the entire database. In the case of the parallel
implementation of Oracle, all nodes must have access to all disks of the
database, regardless of where those disks are physically attached.
Node X
Node Y
Application
Application
Cache
LVM
Cache
LVM
VSD
VSD
IP
IP
Disk DD
Net DD
Net DD
Disk DD
IP Network (SP Switch)
lv_X
lv_Y
Figure 17. VSD Architecture
190
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
running the same application. The nodes are connected by the switch and
have locally-attached disks. On Node X’s disk resides a volume group
containing the raw logical volume lv_X. Similarly, Node Y has lv_Y. For the
sake of illustration, let us suppose that lv_X and lv_Y together constitute an
Oracle Parallel Server database to which the application on each node makes
I/O requests.
The application on Node X requests a piece of data in the database. After the
node’s Virtual Memory Manager (VMM) determines that the data is not in
memory, it talks not to the regular Logical Volume Manager (LVM), but rather
to the VSD device driver. The VSD device driver is loaded as a kernel
extension. Thus VSDs configured in the SP are known to the appropriate
nodes at the kernel level.
The VSD device driver can fetch the data from one of three places:
1. From the VSD cache, if the data is still there from previous requests. VSD
cache is shared by all VSDs configured on a node. Data is stored in 4KB
blocks, a size optimized for Oracle Parallel Server. If your I/O patterns
involve I/O operations larger than 4KB, we recommend disabling VSD
cache, because its management becomes counterproductive.
2. From lv_X, in which case the VSD device driver exploits Node X’s normal
LVM and Disk Device Driver (Disk DD) pathway to fetch the data.
3. From lv_Y, in which case the VSD device driver issues the request through
the IP and Network Device Driver (Net DD) pathway to access Node Y. For
performance, VSD uses its own stripped-down IP protocol. Once the
request is passed up through Node Y’s Net DD and IP layers, Node Y’s
VSD device driver accesses the data either from VSD cache or from lv_Y.
The VSD server node uses the buddy buffer to temporarily store data for I/O
operations originating at a client node, and to handle requests that are
greater than the IP message size. In contrast to the data in the cache buffer,
the data in a buddy buffer is purged immediately after the I/O operation
completes. Buddy buffers are used only when a shortage in the switch buffer
pool occurs, or, on certain networks with small IP message sizes (for
example, Ethernet). The maximum and minimum size for the buddy buffer
must be defined when the VSD is created. For best performance, you must
ensure that your buddy buffer limits accommodate your I/O transaction sizes
to minimize the packetizing workload of the VSD protocol. Buddy buffers are
discussed in detail in IBM Parallel System Support Programs for AIX
Managing Shared Disks, SA22-7279.
Special RS/6000 SP Topics 191
Download from Www.Somanuals.com. All Manuals Search And Download.
The VSDs in this scenario are mapped to the raw logical volumes lv_X and
lv_Y. Node X is a client of Node Y’s VSD, and vice versa. Node X is also a
direct client of its own VSD (lv_X), and Node Y is a direct client of VSD lv_Y.
VSD configuration is flexible. An interesting property of the architecture is
that a node can be a client of any other node’s VSD(s), with no dependency
on that client node owning a VSD itself. You could set up three nodes with
powerful I/O capacity to be VSD servers, and ten application nodes, with no
disk other than for AIX, PSSP, and the application executables, as clients of
the VSDs on these server nodes.
VSDs are defined in the SDR and managed by either SP SMIT panels or the
VSD Perspective. VSDs can be in one of five states as shown in Figure 18 on
Undefined
define
undefine
VSD information is
available in the SDR
Defined
cfgvsd
cfgvsd
Open/close and I/O
requests fail
Stopped
preparevsd
stopvsd
I/O requests queued and
open/close request serviced
Suspended
resumevsd
suspendvsd
Open/close and
I/O requests serviced
Active
Available
Figure 18. VSD State Transitions
This figure shows the possible states of a VSD and the commands used to
move between states. VSD configuration changes, or manual recovery of a
failed VSD, require you to move the VSD between various states.
The distributed data access aspect of VSD scales well. The SP Switch itself
provides a very high-bandwidth, scalable interconnect between VSD clients
and servers, while the VSD layers of code are efficient. The performance
192
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
impact of servicing a local I/O request through VSD relative to the normal
VMM/LVM pathway is very small. IBM supports any IP network for VSD, but
we recommend the switch for performance.
VSD provides distributed data access, but not a locking mechanism to
preserve data integrity. A separate product such as Oracle Parallel Server
must provide the global locking mechanism.
9.3.2 Recoverable Virtual Shared Disk
Recoverable Virtual Shared Disk (RVSD) adds availability to VSD. RVSD
allows you to twin-tail disks, that is, physically connect the same group of
disks to two or more nodes, and provide transparent failover of VSDs among
the nodes. RVSD is a separately-priced IBM LPP.
Fast IP Network
VSD Server
VSD Server
VSD Client
Node X
Node Y
Node Z
rvsd_X
rvsd_Y
VSD
VSD
VSD
rvsd_X
(lv_X)
rvsd_Y
(lv_Y)
Figure 19. RVSD Function
using VSD. RVSD is installed on Nodes X and Y to protect VSDs rvsd_X and
rvsd_Y. Nodes X and Y physically connect to each other’s disk subsystems
where the VSDs reside. Node X is the primary server for rvsd_X and the
secondary server for rvsd_Y, and vice versa for Node Y. Should Node X fail,
RVSD will automatically fail over rvsd_X to Node Y. Node Y will take
ownership of the disks, varyon the volume group containing rvsd_X and make
the VSD available. Node Y then serves both rvsd_X and rvsd_Y. Any I/O
Special RS/6000 SP Topics 193
Download from Www.Somanuals.com. All Manuals Search And Download.
operation that was in progress, as well as new I/O operations against rvsd_X,
are suspended until failover is complete. When Node X is repaired and
rebooted, RVSD switches the rvsd_X back to its primary, Node X.
daemon controls recovery. It invokes the recovery scripts whenever there is a
change in the group membership, which it is recognizing through the use of
Group Services, which in turn relies on information from Topology Services.
When a failure occurs, the rvsd daemon notifies all surviving providers in the
RVSD node group, so they can begin recovery. Communication adapter
failures are treated the same as node failures.
The hc daemon is also called the Connection Manager. It supports the
development of recoverable applications. The hc daemon maintains a
membership list of the nodes that are currently running hc daemons and an
incarnation number that is changed every time the membership list changes.
The hc daemon shadows the rvsd daemon; recording the same changes in
state and management of VSD that rvsd records. The difference is that hc
only records these changes after rvsd processes them, to assure that RVSD
recovery activities begin and complete before the recovery of hc client
applications takes place. This serialization helps ensure data integrity.
RSVD Daemons
hc
rvsd
Group Services
Topology Services
Adapter
Membership
Node
Membership
Figure 20. RVSD Subsystem and HA Infrastructure
IBM Certification Study Guide AIX HACMP
194
Download from Www.Somanuals.com. All Manuals Search And Download.
9.4 SP Switch as an HACMP Network
One of the fascinating things with an RS/6000 SP is the switch network. It has
developed over time; so, currently there are two types of switches at
customer sites. The “older” HPS or HiPS switch (High Performance Switch),
also known as the TB2 switch, and the “newer” SP Switch, also known as the
TB3 switch.
The HPS switch is no longer supported with PSSP Version 3.1, and the same
applies to HACMP/ES Version 4.3.
The two different types of switches differ in their availability design from the
hardware point of view significantly. For example, any fault service action on
the HPS switch caused a total network disruption for a small fraction of time.
For example, running an Estart to get new nodes up on the switch affected
running nodes.
The SP switch however, was designed to do all actions regarding the switch
fabric on the link level, so only the selected node is affected. All the others
continue working without even noticing that something has happened on the
switch network.
9.4.1 Switch Basics Within HACMP
Although it has already been mentioned in other places, the following is a
short summary of basics you have to remember when you configure a switch
as a network to HACMP.
• As the switch network is a point-to-point network, you must configure it to
HACMP as a private network.
• For IPAT to work on the switch network, you must enable ARP on the
switch network. However, hardware address takeover is not supported
with the switch.
• If you configure IPAT, the service and boot addresses are ifconfig alias
addresses on the css0adapter. Since there is currently no support for
more than one switch adapter in a node, this is the way HACMP covers
the normally-needed second adapter for redundancy.
• The base address for the switch adapter, i.e. the switch address known to
the SDR, should not be configured into an HACMP network. This would
lead to confusion for the PSSP switch management software.
• The netmask associated with the css0 base IP address is used as the
netmask for all HACMP SP Switch network addresses.
Special RS/6000 SP Topics 195
Download from Www.Somanuals.com. All Manuals Search And Download.
9.4.2 Eprimary Management
The SP switch has an internal primary backup concept, where the primary
node, known as the Eprimary, is backed up automatically by a backup node.
So, in case any serious failure happens on the primary, it will resign from
work, and the backup node will take over the switch network handling,
keeping track of routes, working on events, and so on.
HACMP/ES used to have an Eprimary management function with versions
below 4.3; so, if you upgrade to Version 4.3 and also upgrade your switch to
the SP switch, and you had configured Eprimary management previously
within the HACMP definitions, you have to unmanage it.
To check whether the Eprimary is set to be managed, issue the following
command:
odmget -q’name=EPRIMARY’ HACMPsp2
If the switch is set to MANAGE, before changing to the new switch, run the
script:
/usr/es/sbin/cluster/events/utils/cl_HPS_Eprimary unmanage
As the SP switch has its availability concept built-in, there is no need to do it
outside the PSSP software, so, HACMP doesn’t have to take care of it any
more.
9.4.3 Switch Failures
As mentioned before, a node in the SP is still restricted to have a maximum of
one switch adapter installed. Therefore, even with the software being able to
assign a new primary node within the SP and outside of HACMP, the switch
adapter is still a single point of failure.
If the switch adapter in a node resigns from work due to a software or
hardware problem, the switch network is down for that node.
If any application running on that node relies on the switch network, this
means that the application has virtually died on that node. Therefore, it might
be advisable to promote the switch network failure into a node failure, as
Recovery” on page 46. HACMP would be able to recognize the network
failure when you configure the switch network as an HACMP network, and
thus would react with a network_down event, which in turn would shut down
the node from HACMP, causing a takeover.
196
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
In case this node was the Eprimary node on the switch network, and it is an
SP switch, then the RS/6000 SP software would have chosen a new Eprimary
independently from the HACMP software as well.
Special RS/6000 SP Topics 197
Download from Www.Somanuals.com. All Manuals Search And Download.
198
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Chapter 10. HACMP Classic vs. HACMP/ES vs. HANFS
So, why would you prefer to install one version of HACMP instead of
another? This chapter summarizes the differences between them, to give you
an idea in which situation one or the other best matches your needs. The
certification test itself does not refer to these different HACMP flavors, but it is
useful to know the differences anyway.
The following paragraphs are based on the assumption that you are using
Version 4.3. For an overview of previous Versions and their corresponding
10.1 HACMP for AIX Classic
High Availability Cluster Multi-Processing for AIX (HACMP for AIX) Version
4.3 comes in two flavors. One of them directly derives from previous versions,
and therefore is called Classic, and the other, which utilizes another
technology for heartbeating, is called HACMP Extended Scalability
(HACMP/ES) see below for details.
Basically, these two versions differ only in the way the cluster manager keeps
track of the status of nodes, adapters and networks. In the Classic Version,
this is done through the use of Network Interface Modules.
Network Interface Modules (NIMs) monitor the nodes and network
interfaces associated with a cluster. Each
network module monitors one cluster network
using one kind of communication protocol (for
example, Ethernet or FDDI). Each network
module is responsible for maintaining
keepalive traffic with neighboring nodes as
directed by the Cluster Controller, for
providing a link to other nodes on the network
it monitors, and for initiating adapter swaps on
certain networks.
10.2 HACMP for AIX / Enhanced Scalability
HACMP/ES no longer used NIMs, but utilizes a technolgy that was originally
developed on the RS/6000 SPs. Since PSSP Version 2.2. RS/6000 SP
Systems come with the Phoenix technology for managing availability of the
nodes. This technology was already designed as a basic instrument for
© Copyright IBM Corp. 1999
199
Download from Www.Somanuals.com. All Manuals Search And Download.
handling membership and event management by using heartbeats. On the
SP, the original High Availability infrastructure was built on this technology,
and HACMP/ES Version 4.3. is now another instance relying on it. As of AIX
4.3.2 and PSSP 3.1, the High Availability infrastructure, which previously was
tightly coupled to PSSP, was externalized into a package called RISC System
Cluster Technology (RSCT). This package can be installed and run, not only
on SP nodes, but also on regular RS/6000 systems. This allows HACMP/ES
to also be available on non-SP RS/6000s as of Version 4.3.
10.2.1 IBM RISC System Cluster Technology (RSCT)
The High Availability services previously packaged with the IBM PSSP for
AIX Availability Services, also known as the ssp.ha fileset, are now an
integral part of the HACMP/ES software. The IBM RS/6000 Cluster
Technology (RSCT) services provide greater scalability, notify distributed
subsystems of software failure, and coordinate recovery and synchronization
among all subsystems in the software stack.
Packaging these services with HACMP/ES makes it possible to run this
software on all RS/6000s, not just on SP nodes.
RSCT Services include the following components:
Event Manager
A distributed subsystem providing a set of high
availability services. It creates events by matching
information about the state of system resources with
information about resource conditions of interest to
client programs. Client programs, in turn, can use event
notifications to trigger recovery from system failures.
Group Services
A system-wide, fault-tolerant, and highly available
facility for coordinating and monitoring changes to the
state of an application running on a set of nodes. Group
Services helps both in the design and implementation of
fault-tolerant applications and in the consistent recovery
of multiple applications. It accomplishes these two
distinct tasks in an integrated framework.
Topology Service A facility for generating heartbeats over multiple
networks and for providing information about adapter
membership, node membership, and routing. Adapter
and node membership provide indications of adapter
and node failures respectively. Reliable Messaging uses
the routing information to route messages between
nodes around adapter failures.
200
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
See Part 4 of HACMP for AIX, Version 4.3: Enhanced Scalability Installation
and Administration Guide, SC23-4284, for more information on these
services.
10.2.2 Enhanced Cluster Security
With HACMP Version 4.3 comes an option to switch security Mode between
Standard and Enhanced.
Standard
Synchronization is done through the /.rhostsremote command
facilities. To avoid the compromised security that the presence
of this file presents, the administrator is strongly encouraged to
remove these files after the synchronization/verification is done.
Enhanced Kerberos authentication is used for remote commands. That
means the kerberos daemons can decide whether a remote host
is who they claim to be. This is done by granting access on the
basis of tickets, which are provided only to those hosts having
the correct identification.
10.3 High Availability for Network File System for AIX
The HANFS for AIX software provides a reliable NFS server capability by
allowing a backup processor to recover current NFS activity should the
primary NFS server fail.
The HANFS for AIX software supports only two nodes in a cluster.
HANFS for AIX is based on High Availability Cluster Multi-Processing for AIX,
Version 4.3 (HACMP for AIX Classic) product architecture, which ensures
that critical resources, configured as part of a cluster, are highly available for
processing. The HANFS for AIX software extends HACMP for AIX by taking
advantage of AIX extensions to the standard NFS functionality that enable it
to handle duplicate requests correctly and restore lock state during NFS
server failover and reintegration.
Note
A cluster cannot be mixed, that is, have some nodes running the HANFS
for AIX software and other nodes running the HACMP for AIX software. A
single cluster must either have all nodes running the HANFS for AIX
software or all nodes running the HACMP for AIX software. Distinct HANFS
and HACMP clusters, however, are allowed on the same physical network.
HACMP Classic vs. HACMP/ES vs. HANFS 201
Download from Www.Somanuals.com. All Manuals Search And Download.
10.4 Similarities and Differences
All three products have the basic structure in common. They all use the same
concepts and structures. So, a cluster or a network, in the HACMP context, is
the same, no matter what product is being used. There is always a Cluster
Manager controlling the node, keeping track of the cluster’s status, and
triggering events. The differences are in the technologies being used
underneath, or in some special cases, the features available.
The technique of keeping track of the status of a cluster by sending and
receiving heartbeat messages is the major difference between HACMP
Classic and HACMP/ES Version 4.3. HACMP Classic uses the network
modules (NIMs) for this purpose. These communicate their results straight
through to the HACMP Cluster Manager. HACMP/ES, uses the facilities of
RSCT, namely Topology Services, Group Services, and Event Management,
for its heartbeating. Since Version 4.3, the restriction to run HACMP/ES on
RS/6000 SP systems only has been withdrawn. However, if you run it on an
RS/6000 SP, you need to have PSSP Version 3.1 installed. As the HPS
Switch is no longer supported with PSSP Version 3.1, you need to upgrade to
the SP Switch, in case you haven’t already, or you will have a switchless
system.
You can still run HACMP Classic on RS/6000 SP Nodes, just as on
standalone RISC System/6000s. It has no references into the PSSP code
whatsoever.
HANFS for AIX Version 4.3 is basically, a modified HACMP Classic,
enhanced with the capability of the takeover node to recover current NFS
activity, should the primary NFS server fail. By means of AIX extensions to
standard NFS functionality, HANFS for AIX is enabled to handle duplicate
requests correctly or restore the lock state in case of an NFS server failover
or reintegration. Remember though, that HANFS is somewhat restricted, in
that it only supports two-node clusters and cascading resource groups.
10.5 Decision Criteria
Your decision of what type of high availability software you are going to use
can be based on various criteria. Existing hardware is one of them.
If you still use the “old” HPS Switch and don’t want to lose its functionality,
you are bound to the use of PSSP 2.4 or lower. Therefore HACMP Classic or
HACMP/ES up to Version 4.2.2 only is the choice for you.
202
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
For switchless RS/6000 SP systems or SPs with the newer SP Switch, the
decision will be based on a more functional level.
Event Management is much more flexible in HACMP/ES, since you can
define custom events. These events can act on anything that haemdcan
detect, which is virtually anything measurable on an AIX system. How to
customize events is explained in great detail in the redbook HACMP
Enhanced Scalability, SG24-2081.
If you have an NFS server that you need to make highly available, especially
if it is heavily used, and NFS file locking is a major issue, you will need to run
HANFS for AIX.
HACMP Classic vs. HACMP/ES vs. HANFS 203
Download from Www.Somanuals.com. All Manuals Search And Download.
204
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Appendix A. Special Notices
This publication is intended to help System Administrators, System Engineers
and other System Professionals to pass the IBM HACMP Certification Exam.
The information in this publication is not intended as the specification for any
of the following programming interfaces: HACMP, HACMP/ES, HANFS or
HACWS. See the PUBLICATIONS section of the IBM Programming
Announcement for those products for more information about what
publications are considered to be product documentation.
References in this publication to IBM products, programs or services do not
imply that IBM intends to make these available in all countries in which IBM
operates. Any reference to an IBM product, program, or service is not
intended to state or imply that only IBM’s product, program, or service may be
used. Any functionally equivalent program that does not infringe any of IBM’s
intellectual property rights may be used instead of the IBM product, program
or service.
Information in this book was developed in conjunction with use of the
equipment specified, and is limited in application to those specific hardware
and software products and levels.
IBM may have patents or pending patent applications covering subject matter
in this document. The furnishing of this document does not give you any
license to these patents. You can send license inquiries, in writing, to the IBM
Director of Licensing, IBM Corporation, 500 Columbus Avenue, Thornwood,
NY 10594 USA.
Licensees of this program who wish to have information about it for the
purpose of enabling: (i) the exchange of information between independently
created programs and other programs (including this one) and (ii) the mutual
use of the information which has been exchanged, should contact IBM
Corporation, Dept. 600A, Mail Drop 1329, Somers, NY 10589 USA.
Such information may be available, subject to appropriate terms and
conditions, including in some cases, payment of a fee.
The information contained in this document has not been submitted to any
formal IBM test and is distributed AS IS. The information about non-IBM
(“vendor”) products in this manual has been supplied by the vendor and IBM
assumes no responsibility for its accuracy or completeness. The use of this
information or the implementation of any of these techniques is a customer
responsibility and depends on the customer's ability to evaluate and integrate
them into the customer's operational environment. While each item may have
© Copyright IBM Corp. 1999
205
Download from Www.Somanuals.com. All Manuals Search And Download.
been reviewed by IBM for accuracy in a specific situation, there is no
guarantee that the same or similar results will be obtained elsewhere.
Customers attempting to adapt these techniques to their own environments
do so at their own risk.
Any pointers in this publication to external Web sites are provided for
convenience only and do not in any manner serve as an endorsement of
these Web sites.
Any performance data contained in this document was determined in a
controlled environment, and therefore, the results that may be obtained in
other operating environments may vary significantly. Users of this document
should verify the applicable data for their specific environment.
The following document contains examples of data and reports used in daily
business operations. To illustrate them as completely as possible, the
examples contain the names of individuals, companies, brands, and
products. All of these names are fictitious and any similarity to the names and
addresses used by an actual business enterprise is entirely coincidental.
Reference to PTF numbers that have not been released through the normal
distribution process does not imply general availability. The purpose of
including these reference numbers is to alert IBM customers to specific
information relative to the implementation of the PTF when it becomes
available to each customer according to the normal IBM PTF distribution
process.
The following terms are trademarks of the International Business Machines
Corporation in the United States and/or other countries:
AIX
Application System/400
AS/400
AT
BookManager
HACMP/6000
IBM
NetView
POWERserver
RS/6000
SP1
CT
Home Director
Micro Channel
POWERparallel
RISC System/6000
SP
System/390
XT
Ultrastar
400
The following terms are trademarks of other companies:
C-bus is a trademark of Corollary, Inc.
206
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Java and HotJava are trademarks of Sun Microsystems, Incorporated.
Microsoft, Windows, Windows NT, and the Windows 95 logo are trademarks
or registered trademarks of Microsoft Corporation.
PC Direct is a trademark of Ziff Communications Company and is used
by IBM Corporation under license.
Pentium, MMX, ProShare, LANDesk, and ActionMedia are trademarks or
registered trademarks of Intel Corporation in the U.S. and other
countries.
Network File System and NFS are trademarks of SUN Microsystems, Inc.
SUN Microsystems is a trademark of SUN Microsystems, Inc.
UNIX is a registered trademark in the United States and other
countries licensed exclusively through X/Open Company Limited.
Other company, product, and service names may be trademarks or
service marks of others.
Special Notices 207
Download from Www.Somanuals.com. All Manuals Search And Download.
208
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Appendix B. Related Publications
The publications listed in this section are considered particularly suitable for a
more detailed discussion of the topics covered in this redbook.
B.1 International Technical Support Organization Publications
For information on ordering these ITSO publications see “How to Get ITSO
• A Practical Guide to Serial Storage Architecture for AIX, SG24-4599
• HACMP Enhanced Scalability, SG24-2081
• HACMP Enhanced Scalability Handbook, SG24-5328
• HACMP Enhanced Scalability: User-Defined Events, SG24-5327
• HACMP/6000 Customization Examples, SG24-4498
• High Availability on the RISC System/6000 Family, SG24-4551
• Inside the RS/6000 SP, SG24-5145
• Monitoring and Managing IBM SSA Disk Subsystems, SG24-5251
• AIX Version 4.3 Migration Guide, SG24-5116
B.2 Redbooks on CD-ROMs
Redbooks are also available on CD-ROMs. Order a subscription and
receive updates 2-4 times a year at significant savings.
CD-ROM Title
Subscription
Number
Collection Kit
Number
System/390 Redbooks Collection
SBOF-7201
SBOF-7370
SBOF-7240
SBOF-6899
SBOF-6898
SBOF-7270
SBOF-7230
SBOF-7205
SBOF-8700
SBOF-7290
SK2T-2177
SK2T-6022
SK2T-8038
SK2T-8039
SK2T-8044
SK2T-2849
SK2T-8040
SK2T-8041
SK2T-8043
SK2T-8037
Networking and Systems Management Redbooks Collection
Transaction Processing and Data Management Redbook
Lotus Redbooks Collection
Tivoli Redbooks Collection
AS/400 Redbooks Collection
RS/6000 Redbooks Collection (HTML, BkMgr)
RS/6000 Redbooks Collection (PostScript)
RS/6000 Redbooks Collection (PDF Format)
Application Development Redbooks Collection
© Copyright IBM Corp. 1999
209
Download from Www.Somanuals.com. All Manuals Search And Download.
B.3 Other Publications
These publications are also relevant as additional sources of information:
• IBM RS/6000 SP: Planning, Volume 2, Control Workstation and Software
Environment, GA22-7281
• IBM PSSP for AIX: Installation and Migration Guide, GA22-7347
• IBM PSSP for AIX: Managing Shared Disks, SA22-7279
• Adapters, Devices, and Cable Information for Multiple Bus Systems,
SA38-0516
• Adapters, Devices, and Cable Information for Micro Channel Bus
Systems, SA38-0533
• PCI Adapter Placement Reference, SA38-0538
• AIX Commands Reference, SBOF-1877
• AIX Performance Monitoring and Tuning Guide, SC23-2365
• AIX HACMP for AIX,Version 4.3: Concepts and Facilities, SC23-4276
• AIX HACMP for AIX,Version 4.3: Planning Guide, SC23-4277
• AIX HACMP for AIX,Version 4.3: Installation Guide, SC23-4278
• AIX HACMP for AIX,Version 4.3: Administration Guide, SC23-4279
• AIX HACMP for AIX,Version 4.3: Troubleshooting Guide, SC23-4280
• AIX HACMP for AIX,Version 4.3: Programming Locking Applications,
SC23-4281
• AIX HACMP for AIX,Version 4.3: Programming Client Applications,
SC23-4282
• AIX HACMP for AIX,Version 4.3: HANFS for AIX Installation and
Administration Guide, SC23-4283
• AIX HACMP for AIX,Version 4.3: Enhanced Scalability Installation and
Administration Guide, SC23-4284
• AIX HACMP for AIX,Version 4.3: Master Index and Glossary, SC23-4285
210
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
How to Get ITSO Redbooks
This section explains how both customers and IBM employees can find out about ITSO redbooks,
CD-ROMs, workshops, and residencies. A form for ordering books and CD-ROMs is also provided.
This information was current at the time of publication, but is continually subject to change. The latest
information may be found at http://www.redbooks.ibm.com/.
How IBM Employees Can Get ITSO Redbooks
Employees may request ITSO deliverables (redbooks, BookManager BOOKs, and CD-ROMs) and
information about redbooks, workshops, and residencies in the following ways:
• Redbooks Web Site on the World Wide Web
http://w3.itso.ibm.com/
• PUBORDER – to order hardcopies in the United States
• Tools Disks
To get LIST3820s of redbooks, type one of the following commands:
TOOLCAT REDPRINT
TOOLS SENDTO EHONE4 TOOLS2 REDPRINT GET SG24xxxx PACKAGE
TOOLS SENDTO CANVM2 TOOLS REDPRINT GET SG24xxxx PACKAGE (Canadian users only)
To get BookManager BOOKs of redbooks, type the following command:
TOOLCAT REDBOOKS
To get lists of redbooks, type the following command:
TOOLS SENDTO USDIST MKTTOOLS MKTTOOLS GET ITSOCAT TXT
To register for information on workshops, residencies, and redbooks, type the following command:
TOOLS SENDTO WTSCPOK TOOLS ZDISK GET ITSOREGI 1998
• REDBOOKS Category on INEWS
• Online – send orders to: USIB6FPL at IBMMAIL or DKIBMBSH at IBMMAIL
Redpieces
For information so current it is still in the process of being written, look at "Redpieces" on the
Redbooks Web Site (http://www.redbooks.ibm.com/redpieces.html). Redpieces are redbooks in
progress; not all redbooks become redpieces, and sometimes just a few chapters will be published
this way. The intent is to get the information out much more quickly than the formal publishing process
allows.
© Copyright IBM Corp. 1999
211
Download from Www.Somanuals.com. All Manuals Search And Download.
How Customers Can Get ITSO Redbooks
Customers may request ITSO deliverables (redbooks, BookManager BOOKs, and CD-ROMs) and
information about redbooks, workshops, and residencies in the following ways:
• Online Orders – send orders to:
IBMMAIL
Internet
In United States
In Canada
Outside North America
usib6fpl at ibmmail
caibmbkz at ibmmail
dkibmbsh at ibmmail
• Telephone Orders
United States (toll free)
Canada (toll free)
1-800-879-2755
1-800-IBM-4YOU
Outside North America
(long distance charges apply)
(+45) 4810-1020 - German
(+45) 4810-1620 - Italian
(+45) 4810-1270 - Norwegian
(+45) 4810-1120 - Spanish
(+45) 4810-1170 - Swedish
(+45) 4810-1320 - Danish
(+45) 4810-1420 - Dutch
(+45) 4810-1540 - English
(+45) 4810-1670 - Finnish
(+45) 4810-1220 - French
• Mail Orders – send orders to:
IBM Publications
Publications Customer Support 144-4th Avenue, S.W.
P.O. Box 29570
Raleigh, NC 27626-0570
USA
IBM Publications
IBM Direct Services
Sortemosevej 21
DK-3450 Allerød
Denmark
Calgary, Alberta T2P 3N5
Canada
• Fax – send orders to:
United States (toll free)
Canada
Outside North America
1-800-445-9269
1-800-267-4455
(+45) 48 14 2207 (long distance charge)
• 1-800-IBM-4FAX (United States) or (+1) 408 256 5422 (Outside USA) – ask for:
Index # 4421 Abstracts of new redbooks
Index # 4422 IBM redbooks
Index # 4420 Redbooks for last six months
• On the World Wide Web
Redbooks Web Site
http://www.redbooks.ibm.com
IBM Direct Publications Catalog
http://www.elink.ibmlink.ibm.com/pbl/pbl
Redpieces
For information so current it is still in the process of being written, look at “"Redpieces”" on the
Redbooks Web Site (http://www.redbooks.ibm.com/redpieces.html). Redpieces are redbooks in
progress; not all redbooks become redpieces, and sometimes just a few chapters will be published
this way. The iinntteennttiissttooggeettththeeininfoforrmmaatitoionnoouuttmmuucchhmqourieckqeuricthkalynththaenftohremfaolrmpuabl lpisuhbilnisghpinrgocpersoscess
allows.
212
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
IBM Redbook Order Form
Please send me the following:
Title
Order Number
Quantity
First name
Company
Address
City
Last name
Postal code
Country
Telephone number
Telefax number
VAT number
Invoice to customer number
Credit card number
Credit card expiration date
Card issued to
Signature
We accept American Express, Diners, Eurocard, Master Card, and Visa. Payment by credit card not
available in all countries. Signature mandatory for credit card payment.
213
Download from Www.Somanuals.com. All Manuals Search And Download.
214
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
List of Abbreviations
GODM
GUI
Global Object Data
Manager
AIX
Advanced Interactive
Executive
Graphical User
Interface
APA
All Points Addressable
APAR
Authorized Program
Analysis Report
HACMP
HANFS
HCON
IBM
High Availability Cluster
Multi-Processing
The description of a
problem to be fixed by
IBM defect support.
This fix is delivered in a
PTF (see below).
High Availability
Network File System
Host Connection
Program
ARP
Address Resolution
Protocol
International Business
Machines Corporation
ASCII
American Standard
Code for Information
Interchange
I/O
IP
Input/Output
Interface Protocol
IPL
Initial Program Load
(System Boot)
AS/400
CDF
Application System/400
Cumulative Distribution
Function
ITSO
International Technical
Support Organization
CD-ROM
Compact Disk - Read
Only Memory
JFS
KA
KB
Kb
Journaled File System
Keepalive Packet
kilobyte
CLM
Cluster Lock Manager
CLVM
Concurrent Logical
Volume Manager
kilobit
CPU
CRM
Central Processing Unit
LAN
LU
Local Area Network
Concurrent Resource
Manager
Logical Unit (SNA
definition)
DE
Differential Ended
Data Link Control
Deadman Switch
LUN
LVM
Logical Unit (RAID
definition)
DLC
DMS
DNS
DSMIT
Logical Volume
Manager
Domain Name Service
MAC
MB
Medium Access Control
megabyte
Distributed System
Management Interface
Tool
MIB
Management
Information Base
FDDI
Fiber Distributed Data
Interface
MTBF
Mean Time Between
Failure
F/W
GB
Fast and Wide (SCSI)
Gigabyte
© Copyright IBM Corp. 1999
215
Download from Www.Somanuals.com. All Manuals Search And Download.
NETBIOS
Network Basic
SPOF
Single Point of Failure
Input/Output System
SPX/IPX
Sequenced Package
Exchange/Internetwork
Packet Exchange
NFS
NIM
Network File System
Network Interface
Module (This is the
definition of NIM in the
HACMP context. NIM in
the AIX 4.1 context
stands for Network
Installation Manager).
SRC
System Resource
Controller
SSA
Serial Storage
Architecture
TCP
Transmission Control
Protocol
NIS
Network Information
Service
TCP/IP
Transmission Control
Protocol/Interface
Protocol
NVRAM
Non-Volatile Random
Access Memory
UDP
User Datagram
Protocol
ODM
POST
PTF
Object Data Manager
Power On Self Test
UPS
Uninterruptible Power
Supply
Program Temporary Fix
A fix to a problem
described in an APAR
(see above).
VGDA
VGSA
WAN
Volume Group
Descriptor Area
Volume Group Status
Area
RAID
Redundant Array of
Independent (or
Inexpensive) Disks
Wide Area Network
RISC
SCSI
SLIP
Reduced Instruction
Set Computer
Small Computer
Systems Interface
Serial Line Interface
Protocol
SMIT
SMP
System Management
Interface Tool
Symmetric
Multi-Processor
SMUX
SNA
SNMP (see below)
Multiplexor
Systems Network
Architecture
SNMP
SOCC
Simple Network
Management Protocol
Serial Optical Channel
Converter
216
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
Index
using 179
clfindres 173
clinfo 156
cllockd 155
clsmuxpd 155
clstat 152
Symbols
/.rhosts file
editing 59
/etc/hosts file
clstrmgr 155
cluster nodes
A
abbreviations 215
acronyms 215
synchronizing 111
Cluster Planning
cluster services
starting
stopping
cluster topology
Clverify 111
concurrent access mode
quorum 90
adding
ATM 13
B
CPU Options
cron 58
cross mounting
C-SPOC 165
C
capacity requirements
cascading resource groups
changing
cl_lsuser command
D
daemons
godm 59
DARE 169
defining
using 179
cl_mkuser command
© Copyright IBM Corp. 1999
217
Download from Www.Somanuals.com. All Manuals Search And Download.
heartbeats 11
dual-network 36
I
Importing 86
E
editing
emsvcsd 156
Eprimary 196
Ethernet 13
J
K
Kerberos 187
L
F
FCS 13
FDDI 13
M
mounting
NFS 126
G
grpglsmd 156
N
grpsvcsd 156
name serving
network adapters
H
HACMP/ES
HACWS 183
planning 40
defined 38
HAView 151
218
IBM Certification Study Guide AIX HACMP
Download from Www.Somanuals.com. All Manuals Search And Download.
networks
point-to-point 36
NFS
S
SCSI
NIM 199
NIS 58
single point of failure
SLIP 13
non-concurrent access
quorum 90
SOCC 13
SSA
advantages 25
SSA disk subsystem
starting
stopping
Symmetric Multi-Processor (SMP)
syncd 146
P
password 49
principal 188
PTFs) 174
R
synchronizing
T
RS232 15
RSCT 200
Rules
TaskGuide 167
Ticket 188
RVSD 193
219
Download from Www.Somanuals.com. All Manuals Search And Download.
ITSO Redbook Evaluation
IBM Certification Study Guide AIX HACMP
SG24-5131-00
Your feedback is very important to help us maintain the quality of ITSO redbooks. Please complete
this questionnaire and return it using one of the following methods:
• Use the online evaluation form found at http://www.redbooks.ibm.com
• Fax this form to: USA International Access Code + 1 914 432 8264
Which of the following best describes you?
_ Customer _ Business Partner
_ None of the above
_ Solution Developer
_ IBM employee
Please rate your overall satisfaction with this book using the scale:
(1 = very good, 2 = good, 3 = average, 4 = poor, 5 = very poor)
Overall Satisfaction
__________
Please answer the following questions:
Was this redbook published in time for your needs?
If no, please explain:
Yes___ No___
What other redbooks would you like to see published?
Comments/Suggestions:
(THANK YOU FOR YOUR FEEDBACK!)
© Copyright IBM Corp. 1999
221
Download from Www.Somanuals.com. All Manuals Search And Download.
IBM Certification Study Guide AIX HACMP
SG24-5131-00
Download from Www.Somanuals.com. All Manuals Search And Download.
|