Product SiteDocumentation Site

Pacemaker 1.0

Configuration Explained

An A-Z guide to Pacemaker's Configuration Options

Edition 1

Andrew Beekhof

Red Hat

Legal Notice

Copyright © 2009 Andrew Beekhof.
The text of and illustrations in this document are licensed under a Creative Commons Attribution–Share Alike 3.0 Unported license ("CC-BY-SA"). An explanation of CC-BY-SA is available at http://creativecommons.org/licenses/by-sa/3.0/.
In accordance with CC-BY-SA, if you distribute this document or an adaptation of it, you must provide the URL for the original version.
In addition to the requirements of this license, the following activities are looked upon favorably:
  1. If you are distributing Open Publication works on hardcopy or CD-ROM, you provide email notification to the authors of your intent to redistribute at least thirty days before your manuscript or media freeze, to give the authors time to provide updated documents. This notification should describe modifications, if any, made to the document.
  2. All substantive modifications (including deletions) be either clearly marked up in the document or else described in an attachment to the document.
  3. Finally, while it is not mandatory under this license, it is considered good form to offer a free copy of any hardcopy or CD-ROM expression of the author(s) work.
Abstract
The purpose of this document is to definitively explain the concepts used to configure Pacemaker. To achieve this best, it will focus exclusively on the XML syntax used to configure the CIB.
For those that are allergic to XML, Pacemaker comes with a cluster shell and a Python based GUI exists, however these tools will not be covered at all in this document [1] , precisely because they hide the XML.
Additionally, this document is NOT a step-by-step how-to guide for configuring a specific clustering scenario. Although such guides exist, the purpose of this document is to provide an understanding of the building blocks that can be used to construct any type of Pacemaker cluster.


[1] It is hoped however, that having understood the concepts explained here, that the functionality of these tools will also be more readily understood.


Table of Contents

Preface
1. Document Conventions
1.1. Typographic Conventions
1.2. Pull-quote Conventions
1.3. Notes and Warnings
2. We Need Feedback!
1. Read-Me-First
1.1. The Scope of this Document
1.2. What Is Pacemaker?
1.3. Types of Pacemaker Clusters
1.4. Pacemaker Architecture
1.4.1. Internal Components
2. Configuration Basics
2.1. Configuration Layout
2.2. The Current State of the Cluster
2.3. How Should the Configuration be Updated?
2.4. Quickly Deleting Part of the Configuration
2.5. Updating the Configuration Without Using XML
2.6. Making Configuration Changes in a Sandbox
2.7. Testing Your Configuration Changes
2.8. Do I Need to Update the Configuration on all Cluster Nodes?
3. Cluster Options
3.1. Special Options
3.1.1. Configuration Version
3.1.2. Other Fields
3.1.3. Fields Maintained by the Cluster
3.2. Cluster Options
3.2.1. Available Cluster Options
3.2.2. Querying and Setting Cluster Options
3.2.3. When Options are Listed More Than Once
4. Cluster Nodes
4.1. Defining a Cluster Node
4.2. Describing a Cluster Node
4.3. Adding a New Cluster Node
4.3.1. Corosync
4.3.2. Heartbeat
4.4. Removing a Cluster Node
4.4.1. Corosync
4.4.2. Heartbeat
4.5. Replacing a Cluster Node
4.5.1. Corosync
4.5.2. Heartbeat
5. Cluster Resources
5.1. What is a Cluster Resource
5.2. Supported Resource Classes
5.2.1. Open Cluster Framework
5.2.2. Linux Standard Base
5.2.3. Legacy Heartbeat
5.3. Properties
5.4. Resource Options
5.5. Setting Global Defaults for Resource Options
5.6. Instance Attributes
5.7. Resource Operations
5.7.1. Monitoring Resources for Failure
5.8. Setting Global Defaults for Operations
5.8.1. When Resources Take a Long Time to Start/Stop
5.8.2. Multiple Monitor Operations
5.8.3. Disabling a Monitor Operation
6. Resource Constraints
6.1. Scores
6.1.1. Infinity Math
6.2. Deciding Which Nodes a Resource Can Run On
6.2.1. Options
6.2.2. Asymmetrical "Opt-In" Clusters
6.2.3. Symmetrical "Opt-Out" Clusters
6.2.4. What if Two Nodes Have the Same Score
6.3. Specifying the Order Resources Should Start/Stop In
6.3.1. Mandatory Ordering
6.3.2. Advisory Ordering
6.4. Placing Resources Relative to other Resources
6.4.1. Options
6.4.2. Mandatory Placement
6.4.3. Advisory Placement
6.5. Ordering Sets of Resources
6.6. Collocating Sets of Resources
7. Receiving Notification of Cluster Events
7.1. Configuring Email Notifications
7.2. Configuring SNMP Notifications
8. Rules
8.1. Node Attribute Expressions
8.2. Time/Date Based Expressions
8.2.1. Date Specifications
8.2.2. Durations
8.3. Using Rules to Determine Resource Location
8.3.1. Using score-attribute Instead of score
8.4. Using Rules to Control Resource Options
8.5. Using Rules to Control Cluster Options
8.6. Ensuring Time Based Rules Take Effect
9. Advanced Configuration
9.1. Connecting to the Cluster Configuration from a Remote Machine
9.2. Specifying When Recurring Actions are Performed
9.3. Moving Resources
9.3.1. Manual Intervention
9.3.2. Moving Resources Due to Failure
9.3.3. Moving Resources Due to Connectivity Changes
9.3.4. Resource Migration
9.4. Reusing Rules, Options and Sets of Operations
9.5. Reloading Services After a Definition Change
10. Advanced Resource Types
10.1. Groups - A Syntactic Shortcut
10.1.1. Properties
10.1.2. Options
10.1.3. Using Groups
10.2. Clones - Resources That Should be Active on Multiple Hosts
10.2.1. Properties
10.2.2. Options
10.2.3. Using Clones
10.3. Multi-state - Resources That Have Multiple Modes
10.3.1. Properties
10.3.2. Options
10.3.3. Using Multi-state Resources
11. Protecting Your Data - STONITH
11.1. Why You Need STONITH
11.2. What STONITH Device Should You Use
11.3. Configuring STONITH
11.3.1. Example
12. Status - Here be dragons
12.1. Node Status
12.2. Transient Node Attributes
12.3. Operation History
12.3.1. Simple Example
12.3.2. Complex Resource History Example
A. FAQ
B. More About OCF Resource Agents
B.1. Location of Custom Scripts
B.2. Actions
B.3. How Does the Cluster Interpret the OCF Return Codes?
B.3.1. Exceptions
C. What Changed in 1.0
C.1. New
C.2. Changed
C.3. Removed
D. Installation
D.1. Choosing a Cluster Stack
D.2. Enabling Pacemaker
D.2.1. For Corosync
D.2.2. For Heartbeat
E. Upgrading Cluster Software
E.1. Version Compatibility
E.2. Complete Cluster Shutdown
E.2.1. Procedure
E.3. Rolling (node by node)
E.3.1. Procedure
E.3.2. Version Compatibility
E.3.3. Crossing Compatibility Boundaries
E.4. Disconnect and Reattach
E.4.1. Procedure
E.4.2. Notes
F. Upgrading the Configuration from 0.6
F.1. Preparation
F.2. Perform the upgrade
F.2.1. Upgrade the software
F.2.2. Upgrade the Configuration
F.2.3. Manually Upgrading the Configuration
G. Is This init Script LSB Compatible?
H. Sample Configurations
H.1. An Empty Configuration
H.2. A Simple Configuration
H.3. An Advanced Configuration
I. Further Reading
J. Revision History
Index

List of Figures

1.1. Active/Passive Redundancy
1.2. Shared Failover
1.3. N to N Redundancy
1.4. Conceptual Stack Overview
1.5. The Pacemaker Stack
1.6. Internal Components
2.1. Sample output from crm_mon
2.2. Sample output from crm_mon -n
2.3. Safely using an editor to modify the cluster configuration
2.4. Safely using an editor to modify a subsection of the cluster configuration
2.5. Searching for STONITH related configuration items
2.6. Creating and displaying the active sandbox
2.7. Small Cluster Transition
2.8. Complex Cluster Transition
4.1. The result of using crm_attribute to specify which kernel pcmk-1 is running
6.1. Ordered Set
6.2. Two Sets of Unordered Resources
6.3. Three Resources Sets
6.4. Another Three Resources Sets
9.1. How the cluster translates the pingd constraint
11.1. Obtaining a list of STONITH Parameters
12.1. A bare-bones status entry for a healthy node called cl-virt-1
12.2. Example set of transient node attributes for node "cl-virt-1"
12.3. A record of the apcstonith resource
12.4. A monitor operation performed by the cluster to determine the current state of the apcstonith resource
12.5. Resource history of a pingd clone with multiple jobs

List of Tables

3.1. Configuration Version Properties
3.2. Properties Controling Validation
3.3. Properties Maintained by the Cluster
3.4. Cluster Options
5.1. Properties of a Primitive Resource
5.2. Options for a Primitive Resource
5.3. Properties of an Operation
6.1. Options for Simple Location Constraints
6.2. Properties of an Ordering Constraint
6.3. Properties of a Collocation Constraint
8.1. Properties of a Rule
8.2. Properties of an Expression
8.3. Properties of a Date Expression
8.4. Properties of a Date Spec
9.1. Environment Variables Used to Connect to Remote Instances of the CIB
9.2. Extra top-level CIB options for remote access
9.3. Common Options for a 'ping' Resource
10.1. Properties of a Group Resource
10.2. Properties of a Clone Resource
10.3. Clone specific configuration options
10.4. Environment variables supplied with Clone notify actions
10.5. Properties of a Multi-State Resource
10.6. Multi-state specific resource configuration options
10.7. Additional constraint options relevant to multi-state resources
10.8. Role implications of OCF return codes
10.9. Environment variables supplied with Master notify actions
12.1. Authoritative Sources for State Information
12.2. Node Status Fields
12.3. Contents of an lrm_rsc_op job.
B.1. Required Actions for OCF Agents
B.2. Optional Actions for OCF Agents
B.3. Types of recovery performed by the cluster
B.4. OCF Return Codes and How They are Handled
E.1. Summary of Upgrade Methodologies
E.2. Version Compatibility Table

List of Examples

2.1. An empty configuration
2.2. Using a sandbox to make multiple changes atomically
3.1. An example of the fields set for a cib object
3.2. Deleting an option that is listed twice
4.1. Example cluster node entry
5.1. An example LSB resource
5.2. An example OCF resource
5.3. An example Heartbeat resource
5.4. An LSB resource with cluster options
5.5. An example OCF resource with instance attributes
5.6. Displaying the metadata for the Dummy resource agent template
5.7. An OCF resource with a recurring health check
5.8. An OCF resource with custom timeouts for its implicit actions
5.9. An OCF resource with two recurring health checks performing different levels of checks
5.10. Example of an OCF resource with a disabled health check
6.1. Example set of opt-in location constraints
6.2. Example set of opt-out location constraints
6.3. Example of two resources that prefer two nodes equally
6.4. Example of an optional and mandatory ordering constraint
6.5. An example colocation constraint
6.6. An example anti-colocation constraint
6.7. An example advisory-only colocation constraint
6.8. A chain of ordered resources
6.9. A chain of ordered resources expressed as a set
6.10. A group resource with the equivalent ordering rules
6.11. Ordered sets of unordered resources
6.12. Advanced use of set ordering - Three ordered sets, two of which are internally unordered
6.13. A chain of collocated resources
6.14. The equivalent colocation chain expressed using resource_sets
6.15. A group resource with the equivalent colocation rules
6.16. Using colocation sets to specify a common peer.
6.17. A colocation chain where the members of the middle set have no inter-dependencies and the last has master status.
8.1. True if now is any time in the year 2005
8.2. Equivalent expression.
8.3. 9am-5pm, Mon-Friday
8.4. 9am-6pm, Mon-Friday, or all day saturday
8.5. 9am-5pm or 9pm-12pm, Mon-Friday
8.6. Mondays in March 2005
8.7. A full moon on Friday the 13th
8.8. Prevent myApacheRsc from running on c001n03
8.9. Prevent myApacheRsc from running on c001n03 - expanded version
8.10. A sample nodes section for use with score-attribute
8.11. Defining different resource options based on the node name
8.12. Set resource-stickiness=INFINITY Mon-Fri between 9am and 6pm, and resource-stickiness=0 all other times
9.1. Specifying a Base for Recurring Action Intervals
9.2. An example ping cluster resource, checks node connectivity once every minute
9.3. Don't run on unconnected nodes
9.4. Run only on nodes connected to 3 or more ping nodes (assumes multiplier is set to 1000)
9.5. Prefer the node with the most connected ping nodes
9.6. A more complex example of choosing a location based on connectivity
9.7. Referencing rules from other constraints
9.8. Referencing attributes, options and operations from other resources
9.9. The DRBD Agent's Control logic for Supporting the reload Operation
9.10. The DRBD Agent Advertising Support for the reload Operation
9.11. Parameter that can be changed using reload
10.1. An example group
10.2. How the cluster sees a group resource
10.3. Example constraints involving groups
10.4. An example clone
10.5. Example constraints involving clones
10.6. Monitoring both states of a multi-state resource
10.7. Example constraints involving multi-state resources
10.8. Manually specifying which node should be promoted
11.1. Sample STONITH Resource
D.1. An example Corosync configuration file
D.2. Example options for an IPv6 environment
D.3. Configuration fragment for enabling Pacemaker under Corosync
D.4. Configuration fragment for enabling Pacemaker under Heartbeat
H.1. An empty configuration
H.2. 2 nodes, some cluster options and a resource
H.3. groups and clones with stonith

Preface

1. Document Conventions

This manual uses several conventions to highlight certain words and phrases and draw attention to specific pieces of information.
In PDF and paper editions, this manual uses typefaces drawn from the Liberation Fonts set. The Liberation Fonts set is also used in HTML editions if the set is installed on your system. If not, alternative but equivalent typefaces are displayed. Note: Red Hat Enterprise Linux 5 and later includes the Liberation Fonts set by default.

1.1. Typographic Conventions

Four typographic conventions are used to call attention to specific words and phrases. These conventions, and the circumstances they apply to, are as follows.
Mono-spaced Bold
Used to highlight system input, including shell commands, file names and paths. Also used to highlight keycaps and key combinations. For example:
To see the contents of the file my_next_bestselling_novel in your current working directory, enter the cat my_next_bestselling_novel command at the shell prompt and press Enter to execute the command.
The above includes a file name, a shell command and a keycap, all presented in mono-spaced bold and all distinguishable thanks to context.
Key combinations can be distinguished from keycaps by the hyphen connecting each part of a key combination. For example:
Press Enter to execute the command.
Press Ctrl+Alt+F2 to switch to the first virtual terminal. Press Ctrl+Alt+F1 to return to your X-Windows session.
The first paragraph highlights the particular keycap to press. The second highlights two key combinations (each a set of three keycaps with each set pressed simultaneously).
If source code is discussed, class names, methods, functions, variable names and returned values mentioned within a paragraph will be presented as above, in mono-spaced bold. For example:
File-related classes include filesystem for file systems, file for files, and dir for directories. Each class has its own associated set of permissions.
Proportional Bold
This denotes words or phrases encountered on a system, including application names; dialog box text; labeled buttons; check-box and radio button labels; menu titles and sub-menu titles. For example:
Choose SystemPreferencesMouse from the main menu bar to launch Mouse Preferences. In the Buttons tab, click the Left-handed mouse check box and click Close to switch the primary mouse button from the left to the right (making the mouse suitable for use in the left hand).
To insert a special character into a gedit file, choose ApplicationsAccessoriesCharacter Map from the main menu bar. Next, choose SearchFind… from the Character Map menu bar, type the name of the character in the Search field and click Next. The character you sought will be highlighted in the Character Table. Double-click this highlighted character to place it in the Text to copy field and then click the Copy button. Now switch back to your document and choose EditPaste from the gedit menu bar.
The above text includes application names; system-wide menu names and items; application-specific menu names; and buttons and text found within a GUI interface, all presented in proportional bold and all distinguishable by context.
Mono-spaced Bold Italic or Proportional Bold Italic
Whether mono-spaced bold or proportional bold, the addition of italics indicates replaceable or variable text. Italics denotes text you do not input literally or displayed text that changes depending on circumstance. For example:
To connect to a remote machine using ssh, type ssh username@domain.name at a shell prompt. If the remote machine is example.com and your username on that machine is john, type ssh john@example.com.
The mount -o remount file-system command remounts the named file system. For example, to remount the /home file system, the command is mount -o remount /home.
To see the version of a currently installed package, use the rpm -q package command. It will return a result as follows: package-version-release.
Note the words in bold italics above — username, domain.name, file-system, package, version and release. Each word is a placeholder, either for text you enter when issuing a command or for text displayed by the system.
Aside from standard usage for presenting the title of a work, italics denotes the first use of a new and important term. For example:
Publican is a DocBook publishing system.

1.2. Pull-quote Conventions

Terminal output and source code listings are set off visually from the surrounding text.
Output sent to a terminal is set in mono-spaced roman and presented thus:
books        Desktop   documentation  drafts  mss    photos   stuff  svn
books_tests  Desktop1  downloads      images  notes  scripts  svgs
Source-code listings are also set in mono-spaced roman but add syntax highlighting as follows:
package org.jboss.book.jca.ex1;

import javax.naming.InitialContext;

public class ExClient
{
   public static void main(String args[]) 
       throws Exception
   {
      InitialContext iniCtx = new InitialContext();
      Object         ref    = iniCtx.lookup("EchoBean");
      EchoHome       home   = (EchoHome) ref;
      Echo           echo   = home.create();

      System.out.println("Created Echo");

      System.out.println("Echo.echo('Hello') = " + echo.echo("Hello"));
   }
}

1.3. Notes and Warnings

Finally, we use three visual styles to draw attention to information that might otherwise be overlooked.

Note

Notes are tips, shortcuts or alternative approaches to the task at hand. Ignoring a note should have no negative consequences, but you might miss out on a trick that makes your life easier.

Important

Important boxes detail things that are easily missed: configuration changes that only apply to the current session, or services that need restarting before an update will apply. Ignoring a box labeled 'Important' will not cause data loss but may cause irritation and frustration.

Warning

Warnings should not be ignored. Ignoring warnings will most likely cause data loss.

2. We Need Feedback!

If you find a typographical error in this manual, or if you have thought of a way to make this manual better, we would love to hear from you! Please submit a report in Bugzilla: http://developerbugs.linux-foundation.org/ against the product Pacemaker.
When submitting a bug report, be sure to mention the manual's identifier: Pacemaker_Explained
If you have a suggestion for improving the documentation, try to be as specific as possible when describing it. If you have found an error, please include the section number and some of the surrounding text so we can find it easily.

Chapter 1. Read-Me-First

1.1. The Scope of this Document

The purpose of this document is to definitively explain the concepts used to configure Pacemaker. To achieve this best, it will focus exclusively on the XML syntax used to configure the CIB.
For those that are allergic to XML, Pacemaker comes with a cluster shell and a Python based GUI exists, however these tools will not be covered at all in this document [2] , precisely because they hide the XML.
Additionally, this document is NOT a step-by-step how-to guide for configuring a specific clustering scenario. Although such guides exist, the purpose of this document is to provide an understanding of the building blocks that can be used to construct any type of Pacemaker cluster.

1.2. What Is Pacemaker?

Pacemaker is a cluster resource manager. It achieves maximum availability for your cluster services (aka. resources) by detecting and recovering from node and resource-level failures by making use of the messaging and membership capabilities provided by your preferred cluster infrastructure (either Corosync or Heartbeat).
Pacemaker's key features include:
  • Detection and recovery of node and service-level failures
  • Storage agnostic, no requirement for shared storage
  • Resource agnostic, anything that can be scripted can be clustered
  • Supports STONITH for ensuring data integrity
  • Supports large and small clusters
  • Supports both quorate and resource driven clusters
  • Supports practically any redundancy configuration
  • Automatically replicated configuration that can be updated from any node
  • Ability to specify cluster-wide service ordering, colocation and anti-colocation
  • Support for advanced services type
    • Clones: for services which need to be active on multiple nodes
    • Multi-state: for services with multiple modes (eg. master/slave, primary/secondary)
  • Unified, scriptable, cluster shell

1.3. Types of Pacemaker Clusters

Pacemaker makes no assumptions about your environment, this allows it to support practically any redundancy configuration including Active/Active, Active/Passive, N+1, N+M, N-to-1 and N-to-N.
Active/Passive Redundancy
Two-node Active/Passive clusters using Pacemaker and DRBD are a cost-effective solution for many High Availability situations.
Figure 1.1. Active/Passive Redundancy

Shared Failover
By supporting many nodes, Pacemaker can dramatically reduce hardware costs by allowing several active/passive clusters to be combined and share a common backup node
Figure 1.2. Shared Failover

N to N Redundancy
When shared storage is available, every node can potentially be used for failover. Pacemaker can even run multiple copies of services to spread out the workload.
Figure 1.3. N to N Redundancy

1.4. Pacemaker Architecture

At the highest level, the cluster is made up of three pieces:
  • Core cluster infrastructure providing messaging and membership functionality (illustrated in red)
  • Non-cluster aware components (illustrated in blue). In a Pacemaker cluster, these pieces include not only the scripts that knows how to start, stop and monitor resources, but also a local daemon that masks the differences between the different standards these scripts implement.
  • A brain (illustrated in green) that processes and reacts to events from the cluster (nodes leaving or joining) and resources (eg. monitor failures) as well as configuration changes from the administrator. In response to all of these events, Pacemaker will compute the ideal state of the cluster and plot a path to achieve it. This may include moving resources, stopping nodes and even forcing them offline with remote power switches.
Conceptual Stack Overview
Conceptual overview of the cluster stack
Figure 1.4. Conceptual Stack Overview

When combined with Corosync, Pacemaker also supports popular open source cluster filesystems [3] Due to recent standardization within the cluster filesystem community, they make use of a common distributed lock manager which makes use of Corosync for its messaging capabilities and Pacemaker for its membership (which nodes are up/down) and fencing services.
The Pacemaker Stack
The Pacemaker stack when running on Corosync
Figure 1.5. The Pacemaker Stack

1.4.1. Internal Components

Pacemaker itself is composed of four key components (illustrated below in the same color scheme as the previous diagram):
  • CIB (aka. Cluster Information Base)
  • CRMd (aka. Cluster Resource Management daemon)
  • PEngine (aka. PE or Policy Engine)
  • STONITHd
Internal Components
Subsystems of a Pacemaker cluster running on Corosync
Figure 1.6. Internal Components

The CIB uses XML to represent both the cluster's configuration and current state of all resources in the cluster. The contents of the CIB are automatically kept in sync across the entire cluster and are used by the PEngine to compute the ideal state of the cluster and how it should be achieved.
This list of instructions is then fed to the DC (Designated Co-ordinator). Pacemaker centralizes all cluster decision making by electing one of the CRMd instances to act as a master. Should the elected CRMd process, or the node it is on, fail... a new one is quickly established.
The DC carries out the PEngine's instructions in the required order by passing them to either the LRMd (Local Resource Management daemon) or CRMd peers on other nodes via the cluster messaging infrastructure (which in turn passes them on to their LRMd process).
The peer nodes all report the results of their operations back to the DC and based on the expected and actual results, will either execute any actions that needed to wait for the previous one to complete, or abort processing and ask the PEngine to recalculate the ideal cluster state based on the unexpected results.
In some cases, it may be necessary to power off nodes in order to protect shared data or complete resource recovery. For this Pacemaker comes with STONITHd. STONITH is an acronym for Shoot-The-Other-Node-In-The-Head and is usually implemented with a remote power switch. In Pacemaker, STONITH devices are modeled as resources (and configured in the CIB) to enable them to be easily monitored for failure, however STONITHd takes care of understanding the STONITH topology such that its clients simply request a node be fenced and it does the rest.


[2] It is hoped however, that having understood the concepts explained here, that the functionality of these tools will also be more readily understood.

[3] Even though Pacemaker also supports Heartbeat, the filesystems need to use the stack for messaging and membership and Corosync seems to be what they're standardizing on. Technically it would be possible for them to support Heartbeat as well, however there seems little interest in this.

Chapter 2. Configuration Basics

2.1. Configuration Layout

The cluster is written using XML notation and divided into two main sections; configuration and status.
The status section contains the history of each resource on each node and based on this data, the cluster can construct the complete current state of the cluster. The authoritative source for the status section is the local resource manager (lrmd) process on each cluster node and the cluster will occasionally repopulate the entire section. For this reason it is never written to disk and admin's are advised against modifying it in any way.
The configuration section contains the more traditional information like cluster options, lists of resources and indications of where they should be placed. The configuration section is the primary focus of this document.
The configuration section itself is divided into four parts:
  • Configuration options (called crm_config)
  • Nodes
  • Resources
  • Resource relationships (called constraints)
Example 2.1. An empty configuration

  <cib generated="true" admin_epoch="0" epoch="0" num_updates="0" have-quorum="false">
     <configuration>
        <crm_config/>
        <nodes/>
        <resources/>
        <constraints/>
     </configuration>
     <status/>
  </cib>


2.2. The Current State of the Cluster

Before one starts to configure a cluster, it is worth explaining how to view the finished product. For this purpose we have created the crm_mon utility that will display the current state of an active cluster. It can show the cluster status by node or by resource and can be used in either single-shot or dynamically-updating mode. There are also modes for displaying a list of the operations performed (grouped by node and resource) as well as information about failures.
Using this tool, you can examine the state of the cluster for irregularities and see how it responds when you cause or simulate failures.
Details on all the available options can be obtained using the crm_mon --help command.
  # crm_mon
  ============
  Last updated: Fri Nov 23 15:26:13 2007
  Current DC: sles-3 (2298606a-6a8c-499a-9d25-76242f7006ec)
  3 Nodes configured.
  5 Resources configured.
  ============

  Node: sles-1 (1186dc9a-324d-425a-966e-d757e693dc86): online
      192.168.100.181    (heartbeat::ocf:IPaddr):    Started sles-1
      192.168.100.182    (heartbeat:IPaddr):        Started sles-1
      192.168.100.183    (heartbeat::ocf:IPaddr):    Started sles-1
      rsc_sles-1    (heartbeat::ocf:IPaddr):    Started sles-1
      child_DoFencing:2    (stonith:external/vmware):    Started sles-1
  Node: sles-2 (02fb99a8-e30e-482f-b3ad-0fb3ce27d088): standby
  Node: sles-3 (2298606a-6a8c-499a-9d25-76242f7006ec): online
      rsc_sles-2    (heartbeat::ocf:IPaddr):    Started sles-3
      rsc_sles-3    (heartbeat::ocf:IPaddr):    Started sles-3
      child_DoFencing:0    (stonith:external/vmware):    Started sles-3
Figure 2.1. Sample output from crm_mon

  # crm_mon -n
  ============
  Last updated: Fri Nov 23 15:26:13 2007
  Current DC: sles-3 (2298606a-6a8c-499a-9d25-76242f7006ec)
  3 Nodes configured.
  5 Resources configured.
  ============

  Node: sles-1 (1186dc9a-324d-425a-966e-d757e693dc86): online
  Node: sles-2 (02fb99a8-e30e-482f-b3ad-0fb3ce27d088): standby
  Node: sles-3 (2298606a-6a8c-499a-9d25-76242f7006ec): online

  Resource Group: group-1
    192.168.100.181    (heartbeat::ocf:IPaddr):    Started sles-1
    192.168.100.182    (heartbeat:IPaddr):        Started sles-1
    192.168.100.183    (heartbeat::ocf:IPaddr):    Started sles-1
  rsc_sles-1    (heartbeat::ocf:IPaddr):    Started sles-1
  rsc_sles-2    (heartbeat::ocf:IPaddr):    Started sles-3
  rsc_sles-3    (heartbeat::ocf:IPaddr):    Started sles-3
  Clone Set: DoFencing
    child_DoFencing:0    (stonith:external/vmware):    Started sles-3
    child_DoFencing:1    (stonith:external/vmware):    Stopped
    child_DoFencing:2    (stonith:external/vmware):    Started sles-1
Figure 2.2. Sample output from crm_mon -n

The DC (Designated Controller) node is where all the decisions are made and if the current DC fails a new one is elected from the remaining cluster nodes. The choice of DC is of no significance to an administrator beyond the fact that its logs will generally be more interesting.

2.3. How Should the Configuration be Updated?

There are three basic rules for updating the cluster configuration:
  • Rule 1 - Never edit the cib.xml file manually. Ever. I'm not making this up.
  • Rule 2 - Read Rule 1 again.
  • Rule 3 - The cluster will notice if you ignored rules 1 & 2 and refuse to use the configuration.
Now that it is clear how NOT to update the configuration, we can begin to explain how you should.
The most powerful tool for modifying the configuration is the cibadmin command which talks to a running cluster. With cibadmin, the user can query, add, remove, update or replace any part of the configuration and all changes take effect immediately so there is no need to perform a reload-like operation.
The simplest way of using cibadmin is to us it to save the current configuration to a temporary file, edit that file with your favorite text or XML editor and then upload the revised configuration.
  cibadmin --query > tmp.xml
  vi tmp.xml
  cibadmin --replace --xml-file tmp.xml
Figure 2.3. Safely using an editor to modify the cluster configuration

Some of the better XML editors can make use of a Relax NG schema to help make sure any changes you make are valid. The schema describing the configuration can normally be found in /usr/lib/heartbeat/pacemaker.rng on most systems.
If you only wanted to modify the resources section, you could instead do
  cibadmin --query --obj_type resources > tmp.xml
  vi tmp.xml
  cibadmin --replace --obj_type resources --xml-file tmp.xml
Figure 2.4. Safely using an editor to modify a subsection of the cluster configuration

to avoid modifying any other part of the configuration.

2.4. Quickly Deleting Part of the Configuration

Identify the object you wish to delete. eg.

Next identify the resource's tag name and id (in this case we'll chose primitive and child_DoFencing). Then simply execute:
cibadmin --delete --crm_xml ‘<primitive id="child_DoFencing"/>'

2.5. Updating the Configuration Without Using XML

Some common tasks can also be performed with one of the higher level tools that avoid the need to read or edit XML.
To enable stonith for example, one could run:
crm_attribute --attr-name stonith-enabled --attr-value true
Or to see if somenode is allowed to run resources, there is:
crm_standby --get-value --node-uname somenode
Or to find the current location of my-test-rsc one can use:
crm_resource --locate --resource my-test-rsc

2.6. Making Configuration Changes in a Sandbox

Often it is desirable to preview the effects of a series of changes before updating the configuration atomically. For this purpose we have created crm_shadow which creates a "shadow" copy of the configuration and arranges for all the command line tools to use it.
To begin, simply invoke crm_shadow and give it the name of a configuration to create [4] and be sure to follow the simple on-screen instructions.

Warning

Read the above carefully, failure to do so could result in you destroying the cluster's active configuration
 # crm_shadow --create test
 Setting up shadow instance
 Type Ctrl-D to exit the crm_shadow shell
 shadow[test]: 
 shadow[test] # crm_shadow --which
 test
Figure 2.6. Creating and displaying the active sandbox

From this point on, all cluster commands will automatically use the shadow copy instead of talking to the cluster's active configuration. Once you have finished experimenting, you can either commit the changes, or discard them as shown below. Again, be sure to follow the on-screen instructions carefully.
For a full list of crm_shadow options and commands, invoke it with the --help option.
Example 2.2. Using a sandbox to make multiple changes atomically
 shadow[test] # crm_failcount -G -r rsc_c001n01
  name=fail-count-rsc_c001n01 value=0
 shadow[test] # crm_standby -v on -n c001n02
 shadow[test] # crm_standby -G -n c001n02
 name=c001n02 scope=nodes value=on
 shadow[test] # cibadmin --erase --force
 shadow[test] # cibadmin --query

 <cib cib_feature_revision="1" validate-with="pacemaker-1.0" admin_epoch="0" crm_feature_set="3.0" have-quorum="1" epoch="112"
      dc-uuid="c001n01" num_updates="1" cib-last-written="Fri Jun 27 12:17:10 2008">
    <configuration>
       <crm_config/>
       <nodes/>
       <resources/>
       <constraints/>
    </configuration>
    <status/>
 </cib>

  shadow[test] # crm_shadow --delete test --force
  Now type Ctrl-D to exit the crm_shadow shell
  shadow[test] # exit
  # crm_shadow --which
  No shadow instance provided
  # cibadmin -Q

 <cib cib_feature_revision="1" validate-with="pacemaker-1.0" admin_epoch="0" crm_feature_set="3.0" have-quorum="1" epoch="110"
       dc-uuid="c001n01" num_updates="551">
    <configuration>
       <crm_config>
          <cluster_property_set id="cib-bootstrap-options">
             <nvpair id="cib-bootstrap-1" name="stonith-enabled" value="1"/>
             <nvpair id="cib-bootstrap-2" name="pe-input-series-max" value="30000"/>

Making changes in a sandbox and verifying the real configuration is untouched

2.7. Testing Your Configuration Changes

We saw previously how to make a series of changes to a "shadow" copy of the configuration. Before loading the changes back into the cluster (eg. crm_shadow --commit mytest --force), it is often advisable to simulate the effect of the changes with ptest. Eg.
ptest --live-check -VVVVV --save-graph tmp.graph --save-dotfile tmp.dot
The tool uses the same library as the live cluster to show what it would have done given the supplied input. It's output, in addition to a significant amount of logging, is stored in two files tmp.graph and tmp.dot, both are representations of the same thing -- the cluster's response to your changes. In the graph file is stored the complete transition, containing a list of all the actions, their parameters and their pre-requisites. Because the transition graph is not terribly easy to read, the tool also generates a Graphviz dot-file representing the same information.
Small Cluster Transition
An example transition graph as represented by Graphviz
Figure 2.7. Small Cluster Transition

Interpreting the Graphviz output
  • Arrows indicate ordering dependencies
  • Dashed-arrows indicate dependencies that are not present in the transition graph
  • Actions with a dashed border of any color do not form part of the transition graph
  • Actions with a green border form part of the transition graph
  • Actions with a red border are ones the cluster would like to execute but are unrunnable
  • Actions with a blue border are ones the cluster does not feel need to be executed
  • Actions with orange text are pseudo/pretend actions that the cluster uses to simplify the graph
  • Actions with black text are sent to the LRM
  • Resource actions have text of the form rsc_action_interval node
  • Any action depending on an action with a red border will not be able to execute.
  • Loops are really bad. Please report them to the development team.
In the above example, it appears that a new node, node2, has come online and that the cluster is checking to make sure rsc1, rsc2 and rsc3 are not already running there (Indicated by the *_monitor_0 entries). Once it did that, and assuming the resources were not active there, it would have liked to stop rsc1 and rsc2 on node1 and move them to node2. However, there appears to be some problem and the cluster cannot or is not permitted to perform the stop actions which implies it also cannot perform the start actions. For some reason the cluster does not want to start rsc3 anywhere.
For information on the options supported by ptest, use ptest --help
Complex Cluster Transition
Another, slightly more complex, transition graph that you're not expected to be able to read
Figure 2.8. Complex Cluster Transition

2.8. Do I Need to Update the Configuration on all Cluster Nodes?

No. Any changes are immediately synchronized to the other active members of the cluster.
To reduce bandwidth, the cluster only broadcasts the incremental updates that result from your changes and uses MD5 sums to ensure that each copy is completely consistent.


[4] Shadow copies are identified with a name, making it possible to have more than one

Chapter 3. Cluster Options

3.1. Special Options

The reason for these fields to be placed at the top level instead of with the rest of cluster options is simply a matter of parsing. These options are used by the configuration database which is, by design, mostly ignorant of the content it holds. So the decision was made to place them in an easy to find location.

3.1.1. Configuration Version

When a node joins the cluster, the cluster will perform a check to see who has the best configuration based on the fields below. It then asks the node with the highest (admin_epoch, epoch, num_updates) tuple to replace the configuration on all the nodes - which makes setting them and setting them correctly very important.
Table 3.1. Configuration Version Properties
Field Description
admin_epoch
Never modified by the cluster. Use this to make the configurations on any inactive nodes obsolete.
Never set this value to zero, in such cases the cluster cannot tell the difference between your configuration and the "empty" one used when nothing is found on disk.
epoch Incremented every time the configuration is updated (usually by the admin)
num_updates Incremented every time the configuration or status is updated (usually by the cluster)

3.1.2. Other Fields

Table 3.2. Properties Controling Validation
Field Description
validate-with Determines the type of validation being done on the configuration. If set to "none", the cluster will not verify that updates conform the the DTD (nor reject ones that don't). This option can be useful when operating a mixed version cluster during an upgrade.

3.1.3. Fields Maintained by the Cluster

Table 3.3. Properties Maintained by the Cluster
Field Description
crm-debug-origin Indicates where the last update came from. Informational purposes only.
cib-last-written Indicates when the configuration was last written to disk. Informational purposes only.
dc-uuid Indicates which cluster node is the current leader. Used by the cluster when placing resources and determining the order of some events.
have-quorum Indicates if the cluster has quorum. If false, this may mean that the cluster cannot start resources or fence other nodes. See no-quorum-policy below.

Note that although these fields can be written to by the admin, in most cases the cluster will overwrite any values specified by the admin with the "correct" ones. To change the admin_epoch, for example, one would use:
cibadmin --modify --crm_xml ‘<cib admin_epoch="42"/>'
A complete set of fields will look something like this:
Example 3.1. An example of the fields set for a cib object

 <cib have-quorum="true" validate-with="pacemaker-1.0" admin_epoch="1" epoch="12" num_updates="65"
    dc-uuid="ea7d39f4-3b94-4cfa-ba7a-952956daabee">


3.2. Cluster Options

Cluster options, as you'd expect, control how the cluster behaves when confronted with certain situations.
They are grouped into sets and, in advanced configurations, there may be more than one.[5] For now we will describe the simple case where each option is present at most once.

3.2.1. Available Cluster Options

Table 3.4. Cluster Options
Option Default Description
batch-limit 30 The number of jobs that the TE is allowed to execute in parallel. The "correct" value will depend on the speed and load of your network and cluster nodes.
no-quorum-policy stop What to do when the cluster does not have quorum. Allowed values:
  • ignore - continue all resource management
  • freeze - continue resource management, but don't recover resources from nodes not in the affected partition
  • stop - stop all resources in the affected cluster parition
  • suicide - fence all nodes in the affected cluster partition
symmetric-cluster TRUE Can all resources run on any node by default?
stonith-enabled TRUE
Should failed nodes and nodes with resources that can't be stopped be shot? If you value your data, set up a STONITH device and enable this.
If true, or unset, the cluster will refuse to start resources unless one or more STONITH resources have been configured also.
stonith-action reboot Action to send to STONITH device. Allowed values: reboot, poweroff.
cluster-delay 60s Round trip delay over the network (excluding action execution). The "correct" value will depend on the speed and load of your network and cluster nodes.
stop-orphan-resources TRUE Should deleted resources be stopped
stop-orphan-actions TRUE Should deleted actions be cancelled
start-failure-is-fatal TRUE When set to FALSE, the cluster will instead use the resource's failcount and value for resource-failure-stickiness
pe-error-series-max -1 (all) The number of PE inputs resulting in ERRORs to save. Used when reporting problems.
pe-warn-series-max -1 (all) The number of PE inputs resulting in WARNINGs to save. Used when reporting problems.
pe-input-series-max -1 (all) The number of "normal" PE inputs to save. Used when reporting problems.

You can always obtain an up-to-date list of cluster options, including their default values by running the pengine metadata command.

3.2.2. Querying and Setting Cluster Options

Cluster options can be queried and modified using the crm_attribute tool. To get the current value of cluster-delay, simply use:
crm_attribute --attr-name cluster-delay --get-value
which is more simply written as
crm_attribute --get-value -n cluster-delay
If a value is found, the you'll see a result such as this
  # crm_attribute --get-value -n cluster-delay
 name=cluster-delay value=60s
However if no value is found, the tool will display an error:
 # crm_attribute --get-value -n clusta-deway
 name=clusta-deway value=(null)
 Error performing operation: The object/attribute does not exist
To use a different value, eg. 30s, simply run:
crm_attribute --attr-name cluster-delay --attr-value 30s
To go back to the cluster's default value, you can then delete the value with:
crm_attribute --attr-name cluster-delay --delete-attr

3.2.3. When Options are Listed More Than Once

If you ever see something like the following, it means that the option you're modifying is present more than once.
Example 3.2. Deleting an option that is listed twice
 # crm_attribute --attr-name batch-limit --delete-attr
 Multiple attributes match name=batch-limit in crm_config:
 Value: 50          (set=cib-bootstrap-options, id=cib-bootstrap-options-batch-limit)
 Value: 100         (set=custom, id=custom-batch-limit)
 Please choose from one of the matches above and supply the 'id' with --attr-id

In such cases follow the on-screen instructions to perform the requested action. To determine which value is currently being used by the cluster, please refer to the the section on Chapter 8, Rules.


[5] This will be described later in the section on Chapter 8, Rules where we will show how to have the cluster use different sets of options during working hours (when downtime is usually to be avoided at all costs) than it does during the weekends (when resources can be moved to the their preferred hosts without bothering end users)

Chapter 4. Cluster Nodes

4.1. Defining a Cluster Node

Each node in the cluster will have an entry in the nodes section containing its UUID, uname and type.
Example 4.1. Example cluster node entry
  <node id="1186dc9a-324d-425a-966e-d757e693dc86" uname="pcmk-1" type="normal"/>

In normal circumstances, the admin should let the cluster populate this information automatically from the communications and membership data. However one can use the crm_uuid tool to read an existing UUID or define a value before the cluster starts.

4.2. Describing a Cluster Node

Beyond the basic definition of a node, the administrator can also describe the node's attributes, such as how much RAM, disk, what OS or kernel version it has, perhaps even its physical location. This information can then be used by the cluster when deciding where to place resources. For more information on the use of node attributes, see the section on Chapter 8, Rules.
Node attributes can be specified ahead of time or populated later, when the cluster is running, using crm_attribute.
Below is what the node's definition would look like if the admin ran the command:
  crm_attribute --type nodes --node-uname pcmk-1 --attr-name kernel --attr-value `uname -r` 

  <node uname="pcmk-1" type="normal" id="1186dc9a-324d-425a-966e-d757e693dc86">
   <instance_attributes id="nodes-1186dc9a-324d-425a-966e-d757e693dc86">
     <nvpair id="kernel-1186dc9a-324d-425a-966e-d757e693dc86" name="kernel" value="2.6.16.46-0.4-default"/>
   </instance_attributes>
  </node>

Figure 4.1. The result of using crm_attribute to specify which kernel pcmk-1 is running

A simpler way to determine the current value of an attribute is to use crm_attribute command again:
crm_attribute --type nodes --node-uname pcmk-1 --attr-name kernel --get-value
By specifying --type nodes the admin tells the cluster that this attribute is persistent. There are also transient attributes which are kept in the status section which are "forgotten" whenever the node rejoins the cluster. The cluster uses this area to store a record of how many times a resource has failed on that node but administrators can also read and write to this section by specifying --type status.

4.3. Adding a New Cluster Node

4.3.1. Corosync

Adding a new is as simple as installing Corosync and Pacemaker, and copying /etc/corosync/corosync.conf and /etc/ais/authkey (if it exists) from an existing node. You may need to modify the mcastaddr option to match the new node's IP address.
If a log message containing "Invalid digest" appears from Corosync, the keys are not consistent between the machines.

4.3.2. Heartbeat

Provided you specified autojoin any in ha.cf, adding a new is as simple as installing heartbeat and copying ha.cf and authkeys from an existing node.
If not, then after setting up ha.cf and authkeys, you must use the hb_addnode command before starting the new node.

4.4. Removing a Cluster Node

4.4.1. Corosync

Because the messaging and membership layers are the authoritative source for cluster nodes, deleting them from the CIB is not a reliable solution. First one must arrange for heartbeat to forget about the node (pcmk-1 in the example below).
On the host to be removed:
  1. Find and record the node's Corosync id: crm_node -i
  2. Stop the cluster: /etc/init.d/corosync stop
Next, from one of the remaining active cluster nodes:
  1. Tell the cluster to forget about the removed host: crm_node -R COROSYNC_ID
  2. Only now is it safe to delete the node from the CIB with:
    cibadmin --delete --obj_type nodes --crm_xml '<node uname="pcmk-1"/>'
    cibadmin --delete --obj_type status --crm_xml '<node_state uname="pcmk-1"/>'

4.4.2. Heartbeat

Because the messaging and membership layers are the authoritative source for cluster nodes, deleting them from the CIB is not a reliable solution. First one must arrange for heartbeat to forget about the node (pcmk-1 in the example below). To do this, shut down heartbeat on the node and then, from one of the remaining active cluster nodes, run:
hb_delnode pcmk-1
Only then is it safe to delete the node from the CIB with:
cibadmin --delete --obj_type nodes --crm_xml '<node uname="pcmk-1"/>'
cibadmin --delete --obj_type status --crm_xml '<node_state uname="pcmk-1"/>'

4.5. Replacing a Cluster Node

4.5.1. Corosync

The five-step guide to replacing an existing cluster node:
  1. Make sure the old node is completely stopped
  2. Give the new machine the same hostname and IP address as the old one
  3. Install the cluster software :-)
  4. Copy /etc/corosync/corosync.conf and /etc/ais/authkey (if it exists) to the new node
  5. Start the new cluster node
If a log message containing "Invalid digest" appears from Corosync, the keys are not consistent between the machines.

4.5.2. Heartbeat

The seven-step guide to replacing an existing cluster node:
  1. Make sure the old node is completely stopped
  2. Give the new machine the same hostname as the old one
  3. Go to an active cluster node and look up the UUID for the old node in /var/lib/heartbeat/hostcache
  4. Install the cluster software
  5. Copy ha.cf and authkeys to the new node
  6. On the new node, populate it's UUID using crm_uuid -w and the UUID from step 2
  7. Start the new cluster node

Chapter 5. Cluster Resources

5.1. What is a Cluster Resource

The role of a resource agent is to abstract the service it provides and present a consistent view to the cluster, which allows the cluster to be agnostic about the resources it manages. The cluster doesn't need to understand how the resource works because it relies on the resource agent to do the right thing when given a start, stop or monitor command.
For this reason it is crucial that resource agents are well tested.
Typically resource agents come in the form of shell scripts, however they can be written using any technology (such as C, Python or Perl) that the author is comfortable with.

5.2. Supported Resource Classes

There are three basic classes of agents supported by Pacemaker. In order of encouraged usage they are:

5.2.1. Open Cluster Framework

The OCF Spec (as it relates to resource agents can be found at: http://www.opencf.org/cgi-bin/viewcvs.cgi/specs/ra/resource-agent-api.txt?rev=HEAD) [6] and is basically an extension of the Linux Standard Base conventions for init scripts to
  • support parameters
  • make them self describing, and
  • extensible
OCF specs have strict definitions of what exit codes actions must return [7] The cluster follows these specifications exactly, and exiting with the wrong exit code will cause the cluster to behave in ways you will likely find puzzling and annoying. In particular, the cluster needs to distinguish a completely stopped resource from one which is in some erroneous and indeterminate state.
Parameters are passed to the script as environment variables, with the special prefix OCF_RESKEY_. So, if you need to be given a parameter which the user thinks of as ip it will be passed to the script as OCF_RESKEY_ip. The number and purpose of the parameters is completely arbitrary, however your script should advertise any that it supports using the meta-data command.

5.2.2. Linux Standard Base

LSB resource agents are those found in /etc/init.d. Generally they are provided by the OS/distribution and in order to be used with the cluster, must conform to the LSB Spec.
Many distributions claim LSB compliance but ship with broken init scripts. To see if your init script is LSB-compatible, see the FAQ entry Appendix G, Is This init Script LSB Compatible?. The most common problems are:
  • Not implementing the status operation at all
  • Not observing the correct exit status codes for start/stop/status actions
  • Starting a started resource returns an error (this violates the LSB spec)
  • Stopping a stopped resource returns an error (this violates the LSB spec)

5.2.3. Legacy Heartbeat

Version 1 of Heartbeat came with its own style of resource agents and it is highly likely that many people have written their own agents based on its conventions. To enable administrators to continue to use these agents, they are supported by the new cluster manager.
The OCF class is the most preferred one as it is an industry standard, highly flexible (allowing parameters to be passed to agents in a non-positional manner) and self-describing.
There is also an additional class, STONITH, which is used exclusively for fencing related resources. This is discussed later in Chapter 11, Protecting Your Data - STONITH.

5.3. Properties

These values tell the cluster which script to use for the resource, where to find that script and what standards it conforms to.
Table 5.1. Properties of a Primitive Resource
Field Description
id Your name for the resource
class The standard the script conforms to. Allowed values: heartbeat, lsb, ocf, stonith
type The name of the Resource Agent you wish to use. eg. IPaddr or Filesystem
provider The OCF spec allows multiple vendors to supply the same ResourceAgent. To use the OCF resource agents supplied with Heartbeat, you should specify heartbeat here.

Resource definitions can be queried with the crm_resource tool. For example
crm_resource --resource Email --query-xml
might produce
Example 5.1. An example LSB resource
  <primitive id="Email" class="lsb" type="exim"/>

Note

One of the main drawbacks to LSB resources is that they do not allow any parameters
or, for an OCF resource:
Example 5.2. An example OCF resource

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
     <instance_attributes id="params-public-ip">
        <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
     </instance_attributes>
  </primitive>


or, finally for the equivalent legacy Heartbeat resource:
Example 5.3. An example Heartbeat resource

  <primitive id="Public-IP-legacy" class="heartbeat" type="IPaddr">
     <instance_attributes id="params-public-ip-legacy">
        <nvpair id="public-ip-addr-legacy" name="1" value="1.2.3.4"/>
     </instance_attributes>
  </primitive>


Note

Heartbeat resources take only ordered and unnamed parameters. The supplied name therefor indicates the order in which they are passed to the script. Only single digit values are allowed.

5.4. Resource Options

Options are used by the cluster to decide how your resource should behave and can be easily set using the --meta option of the crm_resource command.
Table 5.2. Options for a Primitive Resource
Field Default Description
priority 0 If not all resources can be active, the cluster will stop lower priority resources in order to keep higher priority ones active.
target-role Started
What state should the cluster attempt to keep this resource in? Allowed values:
  • Stopped - Force the resource to
  • Started - Allow the resource to be started (In the case of multi-state resources, they will not promoted to master)
  • Master - Allow the resource to be started and, if appropriate, promoted
is-managed TRUE Is the cluster allowed to start and stop the resource? Allowed values: true, false
resource-stickiness Inherited How much does the resource prefer to stay where it is? Defaults to the value of resource-stickiness in the rsc_defaults section
migration-threshold 0 (disabled) How many failures should occur for this resource on a node before making the node ineligible to host this resource.
failure-timeout 0 (disabled) How many seconds to wait before acting as if the failure had not occurred (and potentially allowing the resource back to the node on which it failed.
multiple-active stop_start
What should the cluster do if it ever finds the resource active on more than one node. Allowed values:
  • block - mark the resource as unmanaged
  • stop_only - stop all active instances and leave them that way
  • stop_start - stop all active instances and start the resource in one location only

If you performed the following commands on the previous LSB Email resource
	crm_resource --meta --resource Email --set-parameter priority --property-value 100
	crm_resource --meta --resource Email --set-parameter multiple-active --property-value block
the resulting resource definition would be
Example 5.4. An LSB resource with cluster options

  <primitive id="Email" class="lsb" type="exim">
     <meta_attributes id="meta-email">
        <nvpair id="email-priority" name="priority" value="100"/>
        <nvpair id="email-active" name="multiple-active" value="block"/>
     </meta_attributes>
  </primitive>


5.5. Setting Global Defaults for Resource Options

To set a default value for a resource option, simply add it to the rsc_defaults section with crm_attribute. Thus,
crm_attribute --type rsc_defaults --attr-name is-managed --attr-value false
would prevent the cluster from starting or stopping any of the resources in the configuration (unless of course the individual resources were specifically enabled and had is-managed set to true).

5.6. Instance Attributes

The scripts of some resource classes (LSB not being one of them) can be given parameters which determine how they behave and which instance of a service they control.
If your resource agent supports parameters, you can add them with the crm_resource command. For instance
crm_resource --resource Public-IP --set-parameter ip --property-value 1.2.3.4
would create an entry in the resource like this
Example 5.5. An example OCF resource with instance attributes

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
    </instance_attributes>
  </primitive>


For an OCF resource, the result would be an environment variable called OCF_RESKEY_ip with a value of 1.2.3.4
The list of instance attributes supported by an OCF script can be found by calling the resource script with the meta-data command. The output contains an XML description of all the supported attributes, their purpose and default values.
Example 5.6. Displaying the metadata for the Dummy resource agent template
    export OCF_ROOT=/usr/lib/ocf; $OCF_ROOT/resource.d/pacemaker/Dummy meta-data

  <?xml version="1.0"?>
  <!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
  <resource-agent name="Dummy" version="0.9">
    <version>1.0</version>
  
    <longdesc lang="en-US">
      This is a Dummy Resource Agent. It does absolutely nothing except 
      keep track of whether its running or not.
      Its purpose in life is for testing and to serve as a template for RA writers.
    </longdesc>
    <shortdesc lang="en-US">Dummy resource agent</shortdesc>
  
    <parameters>
      <parameter name="state" unique="1">
        <longdesc lang="en-US">
          Location to store the resource state in.
        </longdesc>
        <shortdesc lang="en-US">State file</shortdesc>
        <content type="string" default="/var/run//Dummy-{OCF_RESOURCE_INSTANCE}.state" />
      </parameter>
  
      <parameter name="dummy" unique="0">
        <longdesc lang="en-US"> 
          Dummy attribute that can be changed to cause a reload
        </longdesc>
        <shortdesc lang="en-US">Dummy attribute that can be changed to cause a reload</shortdesc>
        <content type="string" default="blah" />
      </parameter>
    </parameters>
  
    <actions>
      <action name="start"        timeout="90" />
      <action name="stop"         timeout="100" />
      <action name="monitor"      timeout="20" interval="10" depth="0" start-delay="0" />
      <action name="reload"       timeout="90" />
      <action name="migrate_to"   timeout="100" />
      <action name="migrate_from" timeout="90" />
      <action name="meta-data"    timeout="5" />
      <action name="validate-all" timeout="30" />
    </actions>
  </resource-agent>


5.7. Resource Operations

5.7.1. Monitoring Resources for Failure

By default, the cluster will not ensure your resources are still healthy. To instruct the cluster to do this, you need to add a monitor operation to the resource's definition.
Example 5.7. An OCF resource with a recurring health check

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <operations>
     <op id="public-ip-check" name="monitor" interval="60s"/>
    </operations>
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
    </instance_attributes>
  </primitive>


Table 5.3. Properties of an Operation
Field Description
id Your name for the action. Must be unique.
name The action to perform. Common values: monitor, start, stop
interval How frequently (in seconds) to perform the operation. Default value: 0
timeout How long to wait before declaring the action has failed.
requires
What conditions need to be satisfied before this action occurs. Allowed values:
  • nothing - The cluster may start this resource at any time
  • quorum - The cluster can only start this resource if a majority of the configured nodes are active
  • fencing - The cluster can only start this resource if a majority of the configured nodes are active and any failed or unknown nodes have been powered off.
STONITH resources default to nothing, and all others default to fencing if STONITH is enabled and quorum otherwise.
on-fail
The action to take if this action ever fails. Allowed values:
  • ignore - Pretend the resource did not fail
  • block - Don't perform any further operations on the resource
  • stop - Stop the resource and do not start it elsewhere
  • restart - Stop the resource and start it again (possibly on a different node)
  • fence - STONITH the node on which the resource failed
  • standby - Move all resources away from the node on which the resource failed
The default for the stop operation is fence when STONITH is enabled and block otherwise. All other operations default to stop.
enabled If false, the operation is treated as if it does not exist. Allowed values: true, false

5.8. Setting Global Defaults for Operations

To set a default value for a operation option, simply add it to the op_defaults section with crm_attribute. Thus,
crm_attribute --type op_defaults --attr-name timeout --attr-value 20s
would default each operation's timeout to 20 seconds. If an operation's definition also includes a value for timeout, then that value would be used instead (for that operation only).

5.8.1. When Resources Take a Long Time to Start/Stop

There are a number of implicit operations that the cluster will always perform - start, stop and a non-recurring monitor operation (used at startup to check the resource isn't already active). If one of these is taking too long, then you can create an entry for them and simply specify a new value.
Example 5.8. An OCF resource with custom timeouts for its implicit actions

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <operations>
     <op id="public-ip-startup" name="monitor" interval="0" timeout="90s"/>
     <op id="public-ip-start" name="start" interval="0" timeout="180s"/>
     <op id="public-ip-stop" name="stop" interval="0" timeout="15min"/>
    </operations>
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
    </instance_attributes>
  </primitive>


5.8.2. Multiple Monitor Operations

Provided no two operations (for a single resource) have the same name and interval you can have as many monitor operations as you like. In this way you can do a superficial health check every minute and progressively more intense ones at higher intervals.
To tell the resource agent what kind of check to perform, you need to provide each monitor with a different value for a common parameter. The OCF standard creates a special parameter called OCF_CHECK_LEVEL for this purpose and dictates that it is made available to the resource agent without the normal OCF_RESKEY_ prefix.
Whatever name you choose, you can specify it by adding an instance_attributes block to the op tag. Note that it is up to each resource agent to look for the parameter and decide how to use it.
Example 5.9. An OCF resource with two recurring health checks performing different levels of checks

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <operations>
     <op id="public-ip-health-60" name="monitor" interval="60">
       <instance_attributes id="params-public-ip-depth-60">
           <nvpair id="public-ip-depth-60" name="OCF_CHECK_LEVEL" value="10"/>
       </instance_attributes>
     </op>
     <op id="public-ip-health-300" name="monitor" interval="300">
       <instance_attributes id="params-public-ip-depth-300">
           <nvpair id="public-ip-depth-300" name="OCF_CHECK_LEVEL" value="20"/>
       </instance_attributes>
     </op>
    </operations>
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-level" name="ip" value="1.2.3.4"/>
    </instance_attributes>
  </primitive>


5.8.3. Disabling a Monitor Operation

The easiest way to stop a recurring monitor is to just delete it. However there can be times when you only want to disable it temporarily. In such cases, simply add disabled="true" to the operation's definition.
Example 5.10. Example of an OCF resource with a disabled health check

  <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <operations>
     <op id="public-ip-check" name="monitor" interval="60s" disabled="true"/>
    </operations>
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
    </instance_attributes>
  </primitive>


This can be achieved from the command-line by executing
cibadmin -M -X ‘<op id="public-ip-check" disabled="true"/>'
Once you've done whatever you needed to do, you can then re-enable it with
cibadmin -M -X ‘<op id="public-ip-check" disabled="false"/>'


[6] Note: The Pacemaker implementation has been somewhat extended from the OCF Specs, but none of those changes are incompatible with the original OCF specification

[7] Included with the cluster is the ocf-tester script which can be useful in this regard.

Chapter 6. Resource Constraints

6.1. Scores

Scores of all kinds are integral to how the cluster works. Practically everything from moving a resource to deciding which resource to stop in a degraded cluster is achieved by manipulating scores in some way.
Scores are calculated on a per-resource basis and any node with a negative score for a resource can't run that resource. After calculating the scores for a resource, the cluster then chooses the node with the highest one.

6.1.1. Infinity Math

INFINITY is currently defined as 1,000,000 and addition/subtraction with it follows the following 3 basic rules:
  • Any value + INFINITY = INFINITY
  • Any value - INFINITY = -INFINITY
  • INFINITY - INFINITY = -INFINITY

6.2. Deciding Which Nodes a Resource Can Run On

There are two alternative strategies for specifying which nodes a resources can run on. One way is to say that by default they can run anywhere and then create location constraints for nodes that are not allowed. The other option is to have nodes "opt-in"... to start with nothing able to run anywhere and selectively enable allowed nodes.

6.2.1. Options

Table 6.1. Options for Simple Location Constraints
Field Description
id A unique name for the constraint
rsc A resource name
node A node's uname
score Positive values indicate the resource can run on this node. Negative values indicate the resource can not run on this node. Values of +/- INFINITY change "can" to "must".

6.2.2. Asymmetrical "Opt-In" Clusters

To create an opt-in cluster, start by preventing resources from running anywhere by default
crm_attribute --attr-name symmetric-cluster --attr-value false
Then start enabling nodes. The following fragment says that the web server prefers sles-1, the database prefers sles-2 and both can failover to sles-3 if their most preferred node fails.
Example 6.1. Example set of opt-in location constraints

  <constraints>
    <rsc_location id="loc-1" rsc="Webserver" node="sles-1" score="200"/>
    <rsc_location id="loc-2" rsc="Webserver" node="sles-3" score="0"/>
    <rsc_location id="loc-3" rsc="Database" node="sles-2" score="200"/>
    <rsc_location id="loc-4" rsc="Database" node="sles-3" score="0"/>
  </constraints>


6.2.3. Symmetrical "Opt-Out" Clusters

To create an opt-out cluster, start by allowing resources to run anywhere by default
crm_attribute --attr-name symmetric-cluster --attr-value true
Then start disabling nodes. The following fragment is the equivalent of the above opt-in configuration.
Example 6.2. Example set of opt-out location constraints

  <constraints>
    <rsc_location id="loc-1" rsc="Webserver" node="sles-1" score="200"/>
    <rsc_location id="loc-2-dont-run" rsc="Webserver" node="sles-2" score="-INFINITY"/>
    <rsc_location id="loc-3-dont-run" rsc="Database" node="sles-1" score="-INFINITY"/>
    <rsc_location id="loc-4" rsc="Database" node="sles-2" score="200"/>
  </constraints>


Whether you should choose opt-in or opt-out depends both on your personal preference and the make-up of your cluster. If most of your resources can run on most of the nodes, then an opt-out arrangement is likely to result in a simpler configuration. On the other-hand, if most resources can only run on a small subset of nodes an opt-in configuration might be simpler.

6.2.4. What if Two Nodes Have the Same Score

If two nodes have the same score, then the cluster will choose one. This choice may seem random and may not be what was intended, however the cluster was not given enough information to know what was intended.
Example 6.3. Example of two resources that prefer two nodes equally

  <constraints>
    <rsc_location id="loc-1" rsc="Webserver" node="sles-1" score="INFINITY"/>
    <rsc_location id="loc-2" rsc="Webserver" node="sles-2" score="INFINITY"/>
    <rsc_location id="loc-3" rsc="Database" node="sles-1" score="500"/>
    <rsc_location id="loc-4" rsc="Database" node="sles-2" score="300"/>
    <rsc_location id="loc-5" rsc="Database" node="sles-2" score="200"/>
  </constraints>


In the example above, assuming no other constraints and an inactive cluster, Webserver would probably be placed on sles-1 and Database on sles-2. It would likely have placed Webserver based on the node's uname and Database based on the desire to spread the resource load evenly across the cluster. However other factors can also be involved in more complex configurations.

6.3. Specifying the Order Resources Should Start/Stop In

The way to specify the order in which resources should start is by creating rsc_order constraints.
Table 6.2. Properties of an Ordering Constraint
Field Description
id A unique name for the constraint
first The name of a resource that must be started before the then resource is allowed to.
then The name of a resource. This resource will start after the first resource.
score If greater than zero, the constraint is mandatory. Otherwise it is only a suggestion. Default value: INFINITY
symmetrical If true, which is the default, stop the resources in the reverse order. Default value: true

6.3.1. Mandatory Ordering

When the then resource cannot run without the first resource being active, one should use mandatory constraints. To specify a constraint is mandatory, use a scores greater than zero. This will ensure that the then resource will react when the first resource changes state.
  • If the first resource was running and is stopped, the then resource will also be stopped (if it is running)
  • If the first resource was not running and cannot be started, the then resource will be stopped (if it is running)
  • If the first resource is (re)started while the then resource is running, the then resource will be stopped and restarted

6.3.2. Advisory Ordering

On the other-hand, when score="0" is specified for a constraint, the constraint is considered optional and only has an effect when both resources are stopping and or starting. Any change in state by the first resource will have no effect on the then resource.
Example 6.4. Example of an optional and mandatory ordering constraint

  <constraints>
    <rsc_order id="order-1" first="Database" then="Webserver" />
    <rsc_order id="order-2" first="IP" then="Webserver" score="0"/>
  </constraints>


Some additional information on ordering constraints can be found in the document Ordering Explained

6.4. Placing Resources Relative to other Resources

When the location of one resource depends on the location of another one, we call this colocation.
There is an important side-effect of creating a colocation constraint between two resources, that it affects the order in which resources are assigned to a node. If you think about it, its somewhat obvious. You can't place A relative to B unless you know where B is [8]. So when you are creating colocation constraints, it is important to consider whether you should colocate A with B or B with A.
Another thing to keep in mind is that, assuming A is collocated with B, the cluster will also take into account A's preferences when deciding which node to choose for B. For a detailed look at exactly how this occurs, see the Colocation Explained document.

6.4.1. Options

Table 6.3. Properties of a Collocation Constraint
Field Description
id A unique name for the constraint
rsc The colocation source. If the constraint cannot be satisfied, the cluster may decide not to allow the resource to run at all.
with-rsc The colocation target. The cluster will decide where to put this resource first and then decide where to put the resource in the rsc field
score Positive values indicate the resource should run on the same node. Negative values indicate the resources should not run on the same node. Values of +/- INFINITY change "should" to "must".

6.4.2. Mandatory Placement

Mandatory placement occurs any time the constraint's score is +INFINITY or -INFINITY. In such cases, if the constraint can't be satisfied, then the rsc resource is not permitted to run. For score=INFINITY, this includes cases where the with-rsc resource is not active.
If you need resource1 to always run on the same machine as resource2, you would add the following constraint:
Example 6.5. An example colocation constraint
  <rsc_colocation id="colocate" rsc="resource1" with-rsc="resource2" score="INFINITY"/>

Remember, because INFINITY was used, if resource2 can't run on any of the cluster nodes (for whatever reason) then resource1 will not be allowed to run.
Alternatively, you may want the opposite... that resource1 cannot run on the same machine as resource2. In this case use score="-INFINITY"
Example 6.6. An example anti-colocation constraint
  <rsc_colocation id="anti-colocate" rsc="resource1" with-rsc="resource2" score="-INFINITY"/> 

Again, by specifying -INFINTY, the constraint is binding. So if the only place left to run is where resource2 already is, then resource1 may not run anywhere.

6.4.3. Advisory Placement

If mandatory placement is about "must" and "must not", then advisory placement is the "I'd prefer if" alternative. For constraints with scores greater than -INFINITY and less than INFINITY, the cluster will try and accommodate your wishes but may ignore them if the alternative is to stop some of the cluster resources.
Like in life, where if enough people prefer something it effectively becomes mandatory, advisory colocation constraints can combine with other elements of the configuration to behave as if they were mandatory.
Example 6.7. An example advisory-only colocation constraint
  <rsc_colocation id="colocate-maybe" rsc="resource1" with-rsc="resource2" score="500"/>

6.5. Ordering Sets of Resources

A common situation is for an administrator to create a chain of ordered resources, such as:
Example 6.8. A chain of ordered resources

  <constraints>
    <rsc_order id="order-1" first="A" then="B" />
    <rsc_order id="order-2" first="B" then="C" />
    <rsc_order id="order-3" first="C" then="D" />
  </constraints>


Ordered Set
Visual representation of the four resources' start order for the above constraints
Figure 6.1. Ordered Set

To simplify this situation, there is an alternate format for ordering constraints
Example 6.9. A chain of ordered resources expressed as a set

  <constraints>
    <rsc_order id="order-1">
      <resource_set id="ordered-set-example" sequential="true">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
        <resource_ref id="C"/>
        <resource_ref id="D"/>
      </resource_set>
    </rsc_order>
  </constraints>


Note

Resource sets have the same ordering semantics as groups.
Example 6.10. A group resource with the equivalent ordering rules

  <group id="dummy">
    <primitive id="A" .../>
    <primitive id="B" .../>
    <primitive id="C" .../>
    <primitive id="D" .../>
  </group>


While the set-based format is not less verbose, it is significantly easier to get right and maintain. It can also be expanded to allow ordered sets of (un)ordered resources. In the example below, rscA and rscB can both start in parallel, as can rscC and rscD, however rscC and rscD can only start once both rscA and rscB are active.
Example 6.11. Ordered sets of unordered resources

  <constraints>
    <rsc_order id="order-1">
      <resource_set id="ordered-set-1" sequential="false">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
      </resource_set>
      <resource_set id="ordered-set-2" sequential="false">
        <resource_ref id="C"/>
        <resource_ref id="D"/>
      </resource_set>
    </rsc_order>
  </constraints>


Two Sets of Unordered Resources
Visual representation of the start order for two ordered sets of unordered resources
Figure 6.2. Two Sets of Unordered Resources

Of course either or both sets of resources can also be internally ordered (by setting sequential="true") and there is no limit to the number of sets that can be specified.
Example 6.12. Advanced use of set ordering - Three ordered sets, two of which are internally unordered

  <constraints>
    <rsc_order id="order-1">
      <resource_set id="ordered-set-1" sequential="false">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
      </resource_set>
      <resource_set id="ordered-set-2" sequential="true">
        <resource_ref id="C"/>
        <resource_ref id="D"/>
      </resource_set>
      <resource_set id="ordered-set-3" sequential="false">
        <resource_ref id="E"/>
        <resource_ref id="F"/>
      </resource_set>
    </rsc_order>
  </constraints>


Three Resources Sets
Visual representation of the start order for the three sets defined above
Figure 6.3. Three Resources Sets

6.6. Collocating Sets of Resources

Another common situation is for an administrator to create a set of collocated resources. Previously this possible either by defining a resource group (See Section 10.1, “Groups - A Syntactic Shortcut”) which could not always accurately express the design; or by defining each relationship as an individual constraint, causing a constraint explosion as the number of resources and combinations grew.
Example 6.13. A chain of collocated resources

  <constraints>
    <rsc_colocation id="coloc-1" rsc="B" with-rsc="A" score="INFINITY"/>
    <rsc_colocation id="coloc-2" rsc="C" with-rsc="B" score="INFINITY"/>
    <rsc_colocation id="coloc-3" rsc="D" with-rsc="C" score="INFINITY"/>
  </constraints>


To make things easier, we allow an alternate form of colocation constraints using resource_sets. Just like the expanded version, a resource that can't be active also prevents any resource that must be collocated with it from being active. For example if B was not able to run, then both C (and by inference D) must also remain stopped.
Example 6.14. The equivalent colocation chain expressed using resource_sets

  <constraints>
    <rsc_colocation id="coloc-1" score="INFINITY" >
      <resource_set id="collocated-set-example" sequential="true">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
        <resource_ref id="C"/>
        <resource_ref id="D"/>
      </resource_set>
    </rsc_colocation>
  </constraints>


Note

Resource sets have the same colocation semantics as groups.
Example 6.15. A group resource with the equivalent colocation rules

  <group id="dummy">
    <primitive id="A" .../>
    <primitive id="B" .../>
    <primitive id="C" .../>
    <primitive id="D" .../>
  </group>


This notation can also be used in this context to tell the cluster that a set of resources must all be located with a common peer, but have no dependencies on each other. In this scenario, unlike the previous on, B would be allowed to remain active even if A or C (or both) were inactive.
Example 6.16. Using colocation sets to specify a common peer.

  <constraints>
    <rsc_colocation id="coloc-1" score="INFINITY" >
      <resource_set id="collocated-set-1" sequential="false">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
        <resource_ref id="C"/>
      </resource_set>
      <resource_set id="collocated-set-2" sequential="true">
        <resource_ref id="D"/>
      </resource_set>
    </rsc_colocation>
  </constraints>


Of course there is no limit to the number and size of the sets used. The only thing that matters is that in order for any member of set N to be active, all the members of set N+1 must also be active (and naturally on the same node), and that if a set has sequential="true", then in order for member M to be active, member M+1 must also be active. You can even specify the role in which the members of a set must be in using the set's role attribute.
Example 6.17. A colocation chain where the members of the middle set have no inter-dependencies and the last has master status.

  <constraints>
    <rsc_colocation id="coloc-1" score="INFINITY" >
      <resource_set id="collocated-set-1" sequential="true">
        <resource_ref id="A"/>
        <resource_ref id="B"/>
      </resource_set>
      <resource_set id="collocated-set-2" sequential="false">
        <resource_ref id="C"/>
        <resource_ref id="D"/>
        <resource_ref id="E"/>
      </resource_set>
      <resource_set id="collocated-set-2" sequential="true" role="Master">
        <resource_ref id="F"/>
        <resource_ref id="G"/>
      </resource_set>
    </rsc_colocation>
  </constraints>


Another Three Resources Sets
Visual representation of a colocation chain where the members of the middle set have no inter-dependencies
Figure 6.4. Another Three Resources Sets



[8] While the human brain is sophisticated enough to read the constraint in any order and choose the correct one depending on the situation, the cluster is not quite so smart. Yet.

Chapter 7. Receiving Notification of Cluster Events

7.1. Configuring Email Notifications

7.2. Configuring SNMP Notifications

Chapter 8. Rules

Rules can be used to make your configuration more dynamic. One common example is to set one value for resource-stickiness during working hours, to prevent resources from being moved back to their most preferred location, and another on weekends when no-one is around to notice an outage.
Another use of rules might be to assign machines to different processing groups (using a node attribute) based on time and to then use that attribute when creating location constraints.
Each rule can contain a number of expressions, date-expressions and even other rules. The results of the expressions are combined based on the rule's boolean-op field to determine if the rule ultimately evaluates to true or false. What happens next depends on the context in which the rule is being used.
Table 8.1. Properties of a Rule
Field Description
role Limits the rule to only apply when the resource is in that role. Allowed values: Started, Slave, Master. NOTE: A rule with role="Master" can not determine the initial location of a clone instance. It will only affect which of the active instances will be promoted.
score The score to apply if the rule evaluates to "true". Limited to use in rules that are part of location constraints.
score-attribute The node attribute to look up and use as a score if the rule evaluates to "true". Limited to use in rules that are part of location constraints.
boolean-op How to combine the result of multiple expression objects. Allowed values: and, or

8.1. Node Attribute Expressions

Expression objects are used to control a resource based on the attributes defined by a node or nodes. In addition to any attributes added by the administrator, each node has a built-in node attribute called #uname that can also be used.
Table 8.2. Properties of an Expression
Field Description
value User supplied value for comparison
attribute The node attribute to test
type Determines how the value(s) should be tested. Allowed values: integer, string, version
operation
The comparison to perform. Allowed values:
  • lt,- True if the node attribute's value is less than value
  • gt - True if the node attribute's value is greater than value
  • lte- True if the node attribute's value is less than or equal to value
  • gte- True if the node attribute's value is greater than or equal to value
  • eq- True if the node attribute's value is equal to value
  • ne - True if the node attribute's value is not equal to value
  • defined- True if the node has an the named attribute
  • not_defined- True if the node does not have the named attribute

8.2. Time/Date Based Expressions

As the name suggests, date_expressions are used to control a resource or cluster option based on the current date/time. They can contain an optional date_spec and/or duration object depending on the context.
Table 8.3. Properties of a Date Expression
Field Description
start A date/time conforming to the ISO8601 specification.
end A date/time conforming to the ISO8601 specification. Can be inferred by supplying a value for start and a duration.
operation
Compares the current date/time with the start and/or end date, depending on the context. Allowed values:
  • gt - True if the current date/time is after start
  • lt - True if the current date/time is before end
  • in-range - True if the current date/time is after start and before end
  • date-spec - performs a cron-like comparison between the contents of date_spec and now

Note

Because the comparisons (except for date_spec) include the time, the eq, neq, gte and lte operators have not been implemented.

8.2.1. Date Specifications

date_spec objects are used to create cron-like expressions relating to time. Each field can contain a single number or a single range. Instead of defaulting to zero, any field not supplied is ignored.
For example, monthdays="1" matches the first day of every month and hours="09-17" matches the hours between 9am and 5pm inclusive). However at this time one cannot specify weekdays="1,2" or weekdays="1-2,5-6" since they contain multiple ranges. Depending on demand, this may be implemented in a future release.
Table 8.4. Properties of a Date Spec
Field Description
id A unique name for the date
hours Allowed values: 0-23
monthdays Allowed values: 0-31 (depending on current month and year)
weekdays Allowed values: 1-7 (1=Monday, 7=Sunday)
yeardays Allowed values: 1-366 (depending on the current year)
months Allowed values: 1-12
weeks Allowed values: 1-53 (depending on weekyear)
years Year according the Gregorian calendar
weekyears
May differ from Gregorian years.
Eg. "2005-001 Ordinal" is also "2005-01-01 Gregorian" is also "2004-W53-6 Weekly"
moon Allowed values: 0..7 (0 is new, 4 is full moon). Seriously, you can use this. This was implemented to demonstrate the ease with which new comparisons could be added.

8.2.2. Durations

Durations are used to calculate a value for end when one is not supplied to in_range operations. They contain the same fields as date_spec objects but without the limitations (ie. you can have a duration of 19 days). Like date_specs, any field not supplied is ignored.

8.2.2.1. Sample Time Based Expressions

Example 8.1. True if now is any time in the year 2005

  <rule id="rule1">
   <date_expression id="date_expr1" start="2005-001" operation="in_range">
    <duration years="1"/>
   </date_expression>
  </rule>


Example 8.2. Equivalent expression.

  <rule id="rule2">
   <date_expression id="date_expr2" operation="date_spec">
    <date_spec years="2005"/>
   </date_expression>
  </rule>


Example 8.3. 9am-5pm, Mon-Friday

  <rule id="rule3">
   <date_expression id="date_expr3" operation="date_spec">
    <date_spec hours="9-16" days="1-5"/>
   </date_expression>
  </rule>


Example 8.4. 9am-6pm, Mon-Friday, or all day saturday

  <rule id="rule4" boolean_op="or">
   <date_expression id="date_expr4-1" operation="date_spec">
    <date_spec hours="9-16" days="1-5"/>
   </date_expression>
   <date_expression id="date_expr4-2" operation="date_spec">
    <date_spec days="6"/>
   </date_expression>
  </rule>


Example 8.5. 9am-5pm or 9pm-12pm, Mon-Friday

  <rule id="rule5" boolean_op="and">
   <rule id="rule5-nested1" boolean_op="or">
    <date_expression id="date_expr5-1" operation="date_spec">
     <date_spec hours="9-16"/>
    </date_expression>
    <date_expression id="date_expr5-2" operation="date_spec">
     <date_spec hours="21-23"/>
    </date_expression>
   </rule>
   <date_expression id="date_expr5-3" operation="date_spec">
    <date_spec days="1-5"/>
   </date_expression>
  </rule>


Example 8.6. Mondays in March 2005

  <rule id="rule6" boolean_op="and">
   <date_expression id="date_expr6-1" operation="date_spec">
    <date_spec weekdays="1"/>
   </date_expression>
   <date_expression id="date_expr6-2" operation="in_range" start="2005-03-01" end="2005-04-01"/>
  </rule>


NOTE: Because no time is specified, 00:00:00 is implied.
This means that the range includes all of 2005-03-01 but none of 2005-04-01.
You may wish to write end="2005-03-31T23:59:59" to avoid confusion.
Example 8.7. A full moon on Friday the 13th

  <rule id="rule7" boolean_op="and">
   <date_expression id="date_expr7" operation="date_spec">
    <date_spec weekdays="5" monthdays="13" moon="4"/>
   </date_expression>
  </rule>


8.3. Using Rules to Determine Resource Location

If the constraint's outer-most rule evaluates to false, the cluster treats the constraint as if it was not there. When the rule evaluates to true, the node's preference for running the resource is updated with the score associated with the rule.
If this sounds familiar, its because you have been using a simplified syntax for location constraint rules already. Consider the following location constraint:
Example 8.8. Prevent myApacheRsc from running on c001n03
  <rsc_location id="dont-run-apache-on-c001n03" rsc="myApacheRsc" score="-INFINITY" node="c001n03"/>

This constraint can be more verbosely written as:
Example 8.9. Prevent myApacheRsc from running on c001n03 - expanded version

  <rsc_location id="dont-run-apache-on-c001n03" rsc="myApacheRsc">
    <rule id="dont-run-apache-rule" score="-INFINITY">
       <expression id="dont-run-apache-expr" attribute="#uname" operation="eq" value="c00n03"/>
    </rule>
  </rsc_location>


The advantage of using the expanded form is that one can then add extra clauses to the rule, such as limiting the rule such that it only applies during certain times of the day or days of the week (this is discussed in subsequent sections).
It also allows us to match on node properties other than its name. If we rated each machine's CPU power such that the cluster had the following nodes section:
Example 8.10. A sample nodes section for use with score-attribute

   <nodes>
     <node id="uuid1" uname="c001n01" type="normal">
      <instance_attributes id="uuid1-custom_attrs">
        <nvpair id="uuid1-cpu_mips" name="cpu_mips" value="1234"/>
      </instance_attributes>
     </node>
     <node id="uuid2" uname="c001n02" type="normal">
      <instance_attributes id="uuid2-custom_attrs">
        <nvpair id="uuid2-cpu_mips" name="cpu_mips" value="5678"/>
      </instance_attributes>
     </node>
    </nodes>


then we could prevent resources from running on underpowered machines with the rule

  <rule id="need-more-power-rule" score="-INFINITY">
       <expression id=" need-more-power-expr" attribute="cpu_mips" operation="lt" value="3000"/>
  </rule>

8.3.1. Using score-attribute Instead of score

When using score-attribute instead of score, each node matched by the rule has its score adjusted differently, according to its value for the named node attribute. Thus in the previous example, if a rule used score-attribute="cpu_mips", c001n01 would have its preference to run the resource increased by 1234 whereas c001n02 would have its preference increased by 5678.

8.4. Using Rules to Control Resource Options

Often some cluster nodes will be different from their peers, sometimes these differences (the location of a binary or the names of network interfaces) require resources be configured differently depending on the machine they're hosted on.
By defining multiple instance_attributes objects for the resource and adding a rule to each, we can easily handle these special cases.
In the example below, mySpecialRsc will use eth1 and port 9999 when run on node1, eth2 and port 8888 on node2 and default to eth0 and port 9999 for all other nodes.
Example 8.11. Defining different resource options based on the node name

  <primitive id="mySpecialRsc" class="ocf" type="Special" provider="me">
   <instance_attributes id="special-node1" score="3">
    <rule id="node1-special-case" score="INFINITY" >
     <expression id="node1-special-case-expr" attribute="#uname" operation="eq" value="node1"/>
    </rule>
    <nvpair id="node1-interface" name="interface" value="eth1"/>
   </instance_attributes>
   <instance_attributes id="special-node2" score="2" >
    <rule id="node2-special-case" score="INFINITY">
     <expression id="node2-special-case-expr" attribute="#uname" operation="eq" value="node2"/>
    </rule>
    <nvpair id="node2-interface" name="interface" value="eth2"/>
    <nvpair id="node2-port" name="port" value="8888"/>
   </instance_attributes>
   <instance_attributes id="defaults" score="1" >
    <nvpair id="default-interface" name="interface" value="eth0"/>
    <nvpair id="default-port" name="port" value="9999"/>
   </instance_attributes>
  </primitive>


The order in which instance_attributes objects are evaluated is determined by their score (highest to lowest). If not supplied, score defaults to zero and objects with an equal score are processed in listed order. If the instance_attributes object does not have a rule or has a rule that evaluates to true, then for any parameter the resource does not yet have a value for, the resource will use the parameter values defined by the instance_attributes object.

8.5. Using Rules to Control Cluster Options

Controlling cluster options is achieved in much the same manner as specifying different resource options on different nodes.
The difference is that because they are cluster options, one cannot (or should not because they wont work) use attribute based expressions. The following example illustrates how to set a different resource-stickiness value during and outside of work hours. This allows resources to automatically move back to their most preferred hosts, but at a time that (in theory) does not interfere with business activities.
Example 8.12. Set resource-stickiness=INFINITY Mon-Fri between 9am and 6pm, and resource-stickiness=0 all other times

  <rsc_defaults>
   <meta_attributes id="core-hours" score="2">
    <rule id="core-hour-rule" score="0">
      <date_expression id="nine-to-five-Mon-to-Fri" operation="date_spec">
        <date_spec id="nine-to-five-Mon-to-Fri-spec" hours="9-17" weekdays="1-5"/>
      </date_expression>
    </rule>
    <nvpair id="core-stickiness" name="resource-stickiness" value="INFINITY"/>
   </meta_attributes>
   <meta_attributes id="after-hours" score="1" >
    <nvpair id="after-stickiness" name="resource-stickiness" value="0"/>
   </meta_attributes>
  </rsc_defaults>


8.6. Ensuring Time Based Rules Take Effect

A Pacemaker cluster is an event driven system. As such, it wont recalculate the best place for resources to run in unless something (like a resource failure or configuration change) happens. This can mean that a location constraint that only allows resource X to run between 9am and 5pm is not enforced.
If you rely on time based rules, it is essential that you set the cluster-recheck-interval option. This tells the cluster to periodically recalculate the ideal state of the cluster. For example, if you set cluster-recheck-interval=5m, then sometime between 9:00 and 9:05 the cluster would notice that it needs to start resource X, and between 17:00 and 17:05 it would realize it needed to be stopped.
Note that the timing of the actual start and stop actions depends on what else needs to be performed first.

Chapter 9. Advanced Configuration

9.1. Connecting to the Cluster Configuration from a Remote Machine

Provided Pacemaker is installed on a machine, it is possible to connect to the cluster even if the machine itself is not a part of it. To do this, one simply sets up a number of environment variables and runs the same commands as you would when working on a cluster node.
Table 9.1. Environment Variables Used to Connect to Remote Instances of the CIB
Environment Variable Description
CIB_user The user to connect as. Needs to be part of the hacluster group on the target host. Defaults to $USER
CIB_passwd The user's password. Read from the command line if unset
CIB_server The host to contact. Defaults to localhost.
CIB_port The port on which to contact the server. Required.
CIB_encrypted Encrypt network traffic. Defaults to true.

So if c001n01 is an active cluster node and is listening on 1234 for connections, and someguy is a member of the hacluster group. Then the following would prompt for someguy's password and return the cluster's current configuration:
	 export CIB_port=1234; export CIB_server=c001n01; export CIB_user=someguy; cibadmin -Q 
For security reasons, the cluster does not listen remote connections by default. If you wish to allow remote access, you need to set the remote-tls-port (encrypted) or remote-clear-port (unencrypted) top-level options (ie. those kept in the cib tag , like num_updates and epoch).
Table 9.2. Extra top-level CIB options for remote access
Field Description
remote-tls-port Listen for encrypted remote connections on this port. Default: none
remote-clear-port Listen for plaintext remote connections on this port. Default: none

9.2. Specifying When Recurring Actions are Performed

By default, recurring actions are scheduled relative to when the resource started. So if your resource was last started at 14:32 and you have a backup set to be performed every 24 hours, then the backup will always run at in the middle of the business day - hardly desirable.
To specify a date/time that the operation should be relative to, set the operation's interval-origin. The cluster uses this point to calculate the correct start-delay such that the operation will occur at origin + (interval * N).
So if the operation's interval is 24h, it's interval-origin is set to 02:00 and it is currently 14:32, then the cluster would initiate the operation with a start delay of 11 hours and 28 minutes. If the resource is moved to another node before 2am, then the operation is of course cancelled.
The value specified for interval and interval-origin can be any date/time conforming to the ISO8601 standard. By way of example, to specify an operation that would run on the first Monday of 2009 and every Monday after that you would add:
Example 9.1. Specifying a Base for Recurring Action Intervals
  <op id="my-weekly-action" name="custom-action" interval="P7D" interval-origin="2009-W01-1"/>

9.3. Moving Resources

9.3.1. Manual Intervention

There are primarily two occasions when you would want to move a resource from it's current location: when the whole node is under maintenance and when a single resource needs to be moved.
In the case where everything needs to move, since everything eventually comes down to a score, you could create constraints for every resource you have preventing it from running on that node. While the configuration can seem convoluted at times, not even we would require this of administrators.
Instead one can set a special node attribute which tells the cluster "don't let anything run here". There is even a helpful tool to help query and set it called crm_standby. To check the standby status of the current machine, simply run:
crm_standby --get-value
A value of true indicates that the node is NOT able to host any resources and a value of false indicates that it CAN. You can also check the status of other nodes in the cluster by specifying the --node-uname option. Eg.
crm_standby --get-value --node-uname sles-2
To change the current node's standby status, use --attr-value instead of --get-value. Eg.
crm_standby --attr-value
Again, you can change another host's value by supplying a host name with --node-uname.
When only one resource is required to move, we do this by creating location constraints. However once again we provide a user friendly shortcut as part of the crm_resource command which creates and modifies the extra constraints for you. If Email was running on sles-1 and you wanted it moved to a specific location, the command would look something like:
crm_resource -M -r Email -H sles-2
Behind the scenes, the tool will create the following location constraint:
  <rsc_location rsc="Email" node="sles-2" score="INFINITY"/>
It is important to note that subsequent invocations of crm_resource -M are not cumulative. So if you ran:
crm_resource -M -r Email -H sles-2
crm_resource -M -r Email -H sles-3
then it is as if you had never performed the first command.
To allow the resource to move back again, use:
crm_resource -U -r Email
Note the use of the word allow. The resource can move back to its original location but, depending on resource stickiness, it may stay where it is. To be absolutely certain that it moves back to sles-1, move it there before issuing the call to crm_resource -U:
crm_resource -M -r Email -H sles-1
crm_resource -U -r Email
Alternatively, if you only care that the resource should be moved from its current location, try
crm_resource -M -r Email
Which will instead create a negative constraint. Eg.
<rsc_location rsc="Email" node="sles-1" score="-INFINITY"/>
This will achieve the desired effect but will also have long-term consequences. As the tool will warn you, the creation of a -INFINITY constraint will prevent the resource from running on that node until crm_resource -U is used. This includes the situation where every other cluster node is no longer available.
In some cases, such as when resource stickiness is set to INFINITY, it is possible that you will end up with the problem described in Section 6.2.4, “What if Two Nodes Have the Same Score”. The tool can detect some of these cases and deals with them by also creating both a positive and negative constraint. Eg.
Email prefers sles-1 with a score of -INFINITY
Email prefers sles-2 with a score of INFINITY
which has the same long-term consequences as discussed earlier.

9.3.2. Moving Resources Due to Failure

New in 1.0 is the concept of a migration threshold [9]. Simply define migration-threshold=N for a resource and it will migrate to a new node after N failures. There is no threshold defined by default. To determine the resource's current failure status and limits, use crm_mon --failcounts
By default, once the threshold has been reached, node will no longer be allowed to run the failed resource until the administrator manually resets the resource's failcount using crm_failcount (after hopefully first fixing the failure's cause). However it is possible to expire them by setting the resource's failure-timeout option.
So a setting of migration-threshold=2 and failure-timeout=60s would cause the resource to move to a new node after 2 failures and potentially allow it to move back (depending on the stickiness and constraint scores) after one minute.
There are two exceptions to the migration threshold concept and occur when a resource either fails to start or fails to stop. Start failures cause the failcount to be set to INFINITY and thus always cause the resource to move immediately.
Stop failures are slightly different and crucial. If a resource fails to stop and STONITH is enabled, then the cluster will fence the node in order to be able to start the resource elsewhere. If STONITH is not enabled, then the cluster has no way to continue and will not try to start the resource elsewhere, but will try to stop it again after the failure timeout.

Important

Please read Section 8.6, “Ensuring Time Based Rules Take Effect” before enabling this option.

9.3.3. Moving Resources Due to Connectivity Changes

Setting up the cluster to move resources when external connectivity is lost, is a two-step process.

9.3.3.1. Tell Pacemaker to monitor connectivity

To do this, you need to add a ping resource to the cluster. The ping resource uses the system utility of the same name to a test if list of machines (specified by DNS hostname or IPv4/ IPv6 address) are reachable and uses the results to maintain a node attribute normally called pingd. [10]

Note

Older versions of Heartbeat required users to add ping nodes to ha.cf - this is no longer required.

Important

Older versions of Pacemaker used a custom binary called pingd for this functionality, this is now deprecated in favor of ping. If your version of Pacemaker does not contain the ping agent, you can download the latest version from: http://hg.clusterlabs.org/pacemaker/stable-1.0/raw-file/tip/extra/resources/ping
Normally the resource will run on all cluster nodes, which means that you'll need to create a clone. A template for this can be found below along with a description of the most interesting parameters.
Table 9.3. Common Options for a 'ping' Resource
Field Description
dampen The time to wait (dampening) for further changes occur. Use this to prevent a resource from bouncing around the cluster when cluster nodes notice the loss of connectivity at slightly different times.
multiplier The number by which to multiply the number of connected ping nodes by. Useful when there are multiple ping nodes configured.
host_list The machines to contact in order to determine the current connectivity status. Allowed values include resolvable DNS hostnames, IPv4 and IPv6 addresses.

Example 9.2. An example ping cluster resource, checks node connectivity once every minute

  <clone id="Connected">
   <primitive id="ping" provider="pacemaker" class="ocf" type="ping">
    <instance_attributes id="ping-attrs">
      <nvpair id="pingd-dampen" name="dampen" value="5s"/>
      <nvpair id="pingd-multiplier" name="multiplier" value="1000"/>
      <nvpair id="pingd-hosts" name="host_list" value="my.gateway.com www.bigcorp.com"/>
    </instance_attributes>
    <operations>
      <op id="ping-monitor-60s" interval="60s" name="monitor"/>
    </operations>
   </primitive>
  </clone>


Important

You're only half done. The next section deals with telling Pacemaker how to deal with the connectivity status that ocf:pacemaker:ping is recording.

9.3.3.2. Tell Pacemaker how to interpret the connectivity data

NOTE: Before reading the following, please make sure you have read and understood Chapter 8, Rules above.
There are a number of ways to use the connectivity data provided by Heartbeat. The most common setup is for people to have a single ping node and want to prevent the cluster from running a resource on any unconnected node.
Example 9.3. Don't run on unconnected nodes

  <rsc_location id="WebServer-no-connectivity" rsc="Webserver">
   <rule id="ping-exclude-rule" score="-INFINITY" >
    <expression id="ping-exclude" attribute="pingd" operation="not_defined"/>
   </rule>
  </rsc_location>


A more complex setup is to have a number of ping nodes configured. You can require the cluster to only run resources on nodes that can connect to all (or a minimum subset) of them
Example 9.4. Run only on nodes connected to 3 or more ping nodes (assumes multiplier is set to 1000)

  <rsc_location id="WebServer-connectivity" rsc="Webserver">
   <rule id="ping-prefer-rule" score="-INFINITY" >
    <expression id="ping-prefer" attribute="pingd" operation="lt" value="3000"/>
   </rule>
  </rsc_location> 


or instead you can tell the cluster only to prefer nodes with the most connectivity. Just be sure to set the multiplier to a value higher than that of resource-stickiness (and don't set either of them to INFINITY).
Example 9.5. Prefer the node with the most connected ping nodes

  <rsc_location id="WebServer-connectivity" rsc="Webserver">
   <rule id="ping-prefer-rule" score-attribute="pingd" >
    <expression id="ping-prefer" attribute="pingd" operation="defined"/>
   </rule>
  </rsc_location> 


It is perhaps easier to think of this in terms of the simple constraints that the cluster translates it into. For example, if sles-1 is connected to all 5 ping nodes but sles-2 is only connected to 2, then it would be as if you instead had the following constraints in your configuration:

  <rsc_location id="ping-1" rsc="Webserver" node="sles-1" score="5000"/>
  <rsc_location id="ping-2" rsc="Webserver" node="sles-2" score="2000"/>

Figure 9.1. How the cluster translates the pingd constraint

The advantage being that you don't have to manually update them whenever your network connectivity changes.
You can also combine the concepts above into something even more complex. The example below shows how you can prefer the node with the most connected ping nodes provided they have connectivity to at least three (assuming multiplier is set to 1000).
Example 9.6. A more complex example of choosing a location based on connectivity

  <rsc_location id="WebServer-connectivity" rsc="Webserver">
   <rule id="ping-exclude-rule" score="-INFINITY" >
    <expression id="ping-exclude" attribute="pingd" operation="lt" value="3000"/>
   </rule>
   <rule id="ping-prefer-rule" score-attribute="pingd" >
    <expression id="ping-prefer" attribute="pingd" operation="defined"/>
   </rule>
  </rsc_location> 


9.3.4. Resource Migration

Some resources, such as Xen virtual guests, are able to move to another location without lose of state. We call this resource migration and is different from the normal practice of stopping the resource on the first machine and starting it elsewhere.
Not all resources are able to migrate, see the Migration Checklist below, and those that can wont do so in all situations. Conceptually there are two requirements from which the other prerequisites follow:
  • the resource must be active and healthy at the old location
  • everything required for the resource to run must be available on both the old and new locations
The cluster is able to accommodate both push and pull migration models by requiring the resource agent to support two new actions: migrate_to (performed on the current location) and migrate_from (performed on the destination).
In push migration, the process on the current location transfers the to the new location where is it later activated. In this scenario, most of the work would be done in the migrate_to action and, if anything, the activation would occur during migrate_from.
Conversely for pull, the migrate_to action is practically empty and migrate_from does most of the work, extracting the relevant resource state from the old location and activating it.
There is no wrong or right way to implement migration for your service, as long as it works.

9.3.4.1. Migration Checklist

  • The resource may not be a clone.
  • The resource must use an OCF style agent.
  • The resource must not be in a failed or degraded state.
  • The resource must not, directly or indirectly, depend on any primitive or group resources.
  • The resources must support two new actions: migrate_to and migrate_from and advertise them in its metadata.
  • The resource must have the allow-migrate meta-attribute set to true (not the default).
If the resource depends on a clone, and at the time the resource needs to be move, the clone has instances that are stopping and instances that are starting, then the resource will be moved in the traditional manner. The Policy Engine is not yet able to model this situation correctly and so takes the safe (yet less optimal) path.

9.4. Reusing Rules, Options and Sets of Operations

Sometimes a number of constraints need to use the same set of rules and resources need to set the same options an parameters. To simplify this situation, you can refer to an existing object using an id-ref instead of an id.
So if for one resource you have

  <rsc_location id="WebServer-connectivity" rsc="Webserver">
   <rule id="ping-prefer-rule" score-attribute="pingd" >
    <expression id="ping-prefer" attribute="pingd" operation="defined"/>
   </rule>
  </rsc_location>

Then instead of duplicating the rule for all your other resources, you can instead specify
Example 9.7. Referencing rules from other constraints

  <rsc_location id="WebDB-connectivity" rsc="WebDB">
      <rule id-ref="ping-prefer-rule"/>
  </rsc_location> 


Important

The cluster will insist that the rule exists somewhere. Attempting to add a reference to a non-existing rule will cause a validation failure, as will attempting to remove a rule that is referenced elsewhere.
The same principle applies for meta_attributes and instance_attributes as illustrated in the example below
Example 9.8. Referencing attributes, options and operations from other resources

  <primitive id="mySpecialRsc" class="ocf" type="Special" provider="me">
   <instance_attributes id="mySpecialRsc-attrs" score="1" >
    <nvpair id="default-interface" name="interface" value="eth0"/>
    <nvpair id="default-port" name="port" value="9999"/>
   </instance_attributes>
   <meta_attributes id="mySpecialRsc-options">
    <nvpair id="failure-timeout" name="failure-timeout" value="5m"/>
    <nvpair id="migration-threshold" name="migration-threshold" value="1"/>
    <nvpair id="stickiness" name="resource-stickiness" value="0"/>
   </meta_attributes>
   <operations id="health-checks">
     <op id="health-check" name="monitor" interval="60s"/>
     <op id="health-check" name="monitor" interval="30min"/>
    </operations>
  </primitive>
  <primitive id="myOtherlRsc" class="ocf" type="Other" provider="me">
   <instance_attributes id-ref="mySpecialRsc-attrs"/>
   <meta_attributes id-ref="mySpecialRsc-options"/>
   <operations id-ref="health-checks"/>
  </primitive>


9.5. Reloading Services After a Definition Change

The cluster automatically detects changes to the definition of services it manages. However, the normal response is to stop the service (using the old definition) and start it again (with the new definition). This works well, but some services are smart and can be told to use a new set of options without restarting.
To take advantage of this capability, your resource agent must:
  1. Accept the reload operation and perform any required actions.
    The steps required here depend completely on your application
    Example 9.9. The DRBD Agent's Control logic for Supporting the reload Operation
    
      case $1 in
        start)
    	drbd_start
    	;;
        stop)
    	drbd_stop
    	;;
        reload)
    	drbd_reload
    	;;
        monitor)
    	drbd_monitor
    	;;
        *)	
    	drbd_usage
    	exit $OCF_ERR_UNIMPLEMENTED
    	;;
      esac
      exit $?
    
    

  2. Advertise the reload operation in the actions section of its metadata
    Example 9.10. The DRBD Agent Advertising Support for the reload Operation
    
      <?xml version="1.0"?>
      <!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd">
      <resource-agent name="drbd">
        <version>1.1</version>
        
        <longdesc lang="en">
          Master/Slave OCF Resource Agent for DRBD
        </longdesc>
        
        <shortdesc lang="en">
          This resource agent manages a DRBD resource as a master/slave
          resource. DRBD is a shared-nothing replicated storage device.
        </shortdesc>
        
        <parameters>
          <parameter name="drbd_resource" unique="1" required="1">
            <longdesc lang="en">The name of the drbd resource from the drbd.conf file.</longdesc>
            <shortdesc lang="en">drbd resource name</shortdesc>
            <content type="string"/>
          </parameter>
          
          <parameter name="drbdconf" unique="0">
            <longdesc lang="en">Full path to the drbd.conf file.</longdesc>
            <shortdesc lang="en">Path to drbd.conf</shortdesc>
            <content type="string" default="${OCF_RESKEY_drbdconf_default}"/>
          </parameter>
          
        </parameters>
        
        <actions>
          <action name="start"   timeout="240" />
          <action name="reload"  timeout="240" />
          <action name="promote" timeout="90" />
          <action name="demote"  timeout="90" />
          <action name="notify"  timeout="90" />
          <action name="stop"    timeout="100" />
          <action name="meta-data"    timeout="5" />
          <action name="validate-all" timeout="30" />
        </actions>
      </resource-agent>
    
    

  3. Advertise one or more parameters that can take effect using reload.
    Any parameter with the unique set to 0 is eligable to be used in this way.
    Example 9.11. Parameter that can be changed using reload
    
      <parameter name="drbdconf" unique="0">
        <longdesc lang="en">Full path to the drbd.conf file.</longdesc>
        <shortdesc lang="en">Path to drbd.conf</shortdesc>
        <content type="string" default="${OCF_RESKEY_drbdconf_default}"/>
      </parameter>
    
    

Once these requirements are satisfied, the cluster will automatically know to reload, instead of restarting, the resource when a non-unique fields changes.

Note

The metadata is re-read when the resource is started. This may mean that the resource will be restarted the first time, even though you changed a parameter with unique=0

Note

If both a unique and non-unique field is changed simultaneously, the resource will still be restarted.


[9] The naming of this option was unfortunate as it is easily confused with true migration, the process of moving a resource from one node to another without stopping it. Xen virtual guests are the most common example of resources that can be migrated in this manner.

[10] The attribute name is customizable which allows multiple ping groups to be defined

Chapter 10. Advanced Resource Types

10.1. Groups - A Syntactic Shortcut

One of the most common elements of a cluster is a set of resources that need to be located together, start sequentially and stop in the reverse order. To simplify this configuration we support the concept of groups.
Example 10.1. An example group

  <group id="shortcut">
   <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
    <instance_attributes id="params-public-ip">
       <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
    </instance_attributes>
   </primitive>
   <primitive id="Email" class="lsb" type="exim"/>
  </group>


Although the example above contains only two resources, there is no limit to the number of resources a group can contain. The example is also sufficient to explain the fundamental properties of a group:
  • Resources are started in the order they appear in (Public-IP first, then Email)
  • Resources are stopped in the reverse order to which they appear in (Email first, then Public-IP)
  • If a resource in the group can't run anywhere, then nothing after that is allowed to run
    • If Public-IP can’t run anywhere, neither can Email
    • If Email can’t run anywhere, this does not affect Public-IP in any way
The group above is logically equivalent to writing:
Example 10.2. How the cluster sees a group resource

  <configuration>
   <resources>
    <primitive id="Public-IP" class="ocf" type="IPaddr" provider="heartbeat">
     <instance_attributes id="params-public-ip">
        <nvpair id="public-ip-addr" name="ip" value="1.2.3.4"/>
     </instance_attributes>
    </primitive>
    <primitive id="Email" class="lsb" type="exim"/>
   </resources>
   <constraints>
      <rsc_colocation id="xxx" rsc="Email" with-rsc="Public-IP" score="INFINITY"/>
      <rsc_order id="yyy" first="Public-IP" then="Email"/>
   </constraints>
  </configuration>


Obviously as the group grows bigger, the reduced configuration effort can become significant.

10.1.1. Properties

Table 10.1. Properties of a Group Resource
Field Description
id Your name for the group

10.1.2. Options

Options inherited from simple resources: priority, target-role, is-managed

10.1.3. Using Groups

10.1.3.1. Instance Attributes

Groups have no instance attributes, however any that are set here will be inherited by the group's children.

10.1.3.2. Contents

Groups may only contain a collection of primitive cluster resources. To refer to the child of a group resource, just use the child's id instead of the group's.

10.1.3.3. Constraints

Although it is possible to reference the group's children in constraints, it is usually preferable to use the group's name instead.
Example 10.3. Example constraints involving groups

  <constraints>
    <rsc_location id="group-prefers-node1" rsc="shortcut" node="node1" score="500"/>
    <rsc_colocation id="webserver-with-group" rsc="Webserver" with-rsc="shortcut"/>
    <rsc_order id="start-group-then-webserver" first="Webserver" then="shortcut"/>
  </constraints>


10.1.3.4. Stickiness

Stickiness, the measure of how much a resource wants to stay where it is, is additive in groups. Every active member of the group will contribute its stickiness value to the group's total. So if the default resource-stickiness is 100 a group has seven members, five of which are active, then the group as a whole will prefer its current location with a score of 500.

10.2. Clones - Resources That Should be Active on Multiple Hosts

Clones were initially conceived as a convenient way to start N instances of an IP resource and have them distributed throughout the cluster for load balancing. They have turned out to quite useful for a number of purposes including integrating with Red Hat's DLM, the fencing subsystem and OCFS2.
You can clone any resource provided the resource agent supports it.
Three types of cloned resources exist.
  • Anonymous
  • Globally Unique
  • Stateful
Anonymous clones are the simplest type. These resources behave completely identically everywhere they are running. Because of this, there can only be one copy of an anonymous clone active per machine.
Globally unique clones are distinct entities. A copy of the clone running on one machine is not equivalent to another instance on another node. Nor would any two copies on the same node be equivalent.
Example 10.4. An example clone

  <clone id="apache-clone">
    <meta_attributes id="apache-clone-meta">
       <nvpair id="apache-unique" name="globally-unique" value="false"/>
    </meta_attributes>
    <primitive id="apache" class="lsb" type="apache"/>
  </clone>


10.2.1. Properties

Table 10.2. Properties of a Clone Resource
Field Description
id Your name for the clone

10.2.2. Options

Options inherited from simple resources: priority, target-role, is-managed
Table 10.3. Clone specific configuration options
Field Description
clone-max How many copies of the resource to start. Defaults to the number of nodes in the cluster.
clone-node-max How many copies of the resource can be started on a single node. Defaults to 1.
notify When stopping or starting a copy of the clone, tell all the other copies beforehand and when the action was successful. Allowed values: true, false
globally-unique Does each copy of the clone perform a different function? Allowed values: true, false
ordered Should the copies be started in series (instead of in parallel). Allowed values: true, false
interleave Changes the behavior of ordering constraints (between clones/masters) so that instances can start/stop as soon as their peer instance has (rather than waiting for every instance of the other clone has). Allowed values: true, false

10.2.3. Using Clones

10.2.3.1. Instance Attributes

Clones have no instance attributes, however any that are set here will be inherited by the clone's children.

10.2.3.2. Contents

Clones must contain exactly one group or one regular resource.

Warning

You should never reference the name of a clone's child. If you think you need to do this, you probably need to re-evaluate your design.

10.2.3.3. Constraints

In most cases, a clone will have a single copy on each active cluster node. However if this is not the case, you can indicate which nodes the cluster should to preferentially assign copies to with resource location constraints. These constraints are written no differently to those for regular resources except that the clone's id is used.
Ordering constraints behave slightly differently for clones. In the example below, apache-stats will wait until all copies of the clone that need to be started have done so before being started itself. Only if no copies can be started will apache-stats be prevented from being active. Additionally, the clone will wait for apache-stats to be stopped before stopping the clone.
Colocation of a regular (or group) resource with a clone means that the resource can run on any machine with an active copy of the clone. The cluster will choose a copy based on where the clone is running and the rsc resource's own location preferences.
Colocation between clones is also possible. In such cases, the set of allowed locations for the rsc clone is limited to nodes on which the with clone is (or will be) active. Allocation is then performed as-per-normal.
Example 10.5. Example constraints involving clones

  <constraints>
    <rsc_location id="clone-prefers-node1" rsc="apache-clone" node="node1" score="500"/>
    <rsc_colocation id="stats-with-clone" rsc="apache-stats" with="apache-clone"/>
    <rsc_order id="start-clone-then-stats" first="apache-clone" then="apache-stats"/>
  </constraints>


10.2.3.4. Stickiness

To achieve a stable allocation pattern, clones are slightly sticky by default. If no value for resource-stickiness is provided, the clone will use a value of 1. Being a small value, it causes minimal disturbance to the score calculations of other resources but is enough to prevent Pacemaker from needlessly moving copies around the cluster.

10.2.3.5. Resource Agent Requirements

Any resource can be used as an anonymous clone as it requires no additional support from the resource agent. Whether it makes sense to do so depends on your resource and its resource agent.
Globally unique clones do require some additional support in the resource agent. In particular, it must only respond with ${OCF_SUCCESS} if the node has that exact instance active. All other probes for instances of the clone should result in ${OCF_NOT_RUNNING}. Unless of course they are failed, in which case they should return one of the other OCF error codes.
Copies of a clone are identified by appending a colon and a numerical offset. Eg. apache:2
Resource agents can find out how many copies there are by examining the OCF_RESKEY_CRM_meta_clone_max environment variable and which copy it is by examining OCF_RESKEY_CRM_meta_clone.
You should not make any assumptions (based on OCF_RESKEY_CRM_meta_clone) about which copies are active. In particular, the list of active copies will not always be an unbroken sequence, nor always start at 0.

10.2.3.6. Notifications

Supporting notifications requires the notify action to be implemented. Once supported, the notify action will be passed a number of extra variables which, when combined with additional context, can be used to calculate the current state of the cluster and what is about to happen to it.
Table 10.4. Environment variables supplied with Clone notify actions
Variable Description
OCF_RESKEY_CRM_meta_notify_type Allowed values: pre, post
OCF_RESKEY_CRM_meta_notify_operation Allowed values: start, stop
OCF_RESKEY_CRM_meta_notify_start_resource Resources to be started
OCF_RESKEY_CRM_meta_notify_stop_resource Resources to be stopped
OCF_RESKEY_CRM_meta_notify_active_resource Resources the that are running
OCF_RESKEY_CRM_meta_notify_inactive_resource Resources the that are not running
OCF_RESKEY_CRM_meta_notify_start_uname Nodes on which resources will be started
OCF_RESKEY_CRM_meta_notify_stop_uname Nodes on which resources will be stopped
OCF_RESKEY_CRM_meta_notify_active_uname Nodes on which resources are running
OCF_RESKEY_CRM_meta_notify_inactive_uname Nodes on which resources are not running

The variables come in pairs, such as OCF_RESKEY_CRM_meta_notify_start_resource and OCF_RESKEY_CRM_meta_notify_start_uname and should be treated as an array of whitespace separated elements.
Thus in order to indicate that clone:0 will be started on sles-1, clone:2 will be started on sles-3, and clone:3 will be started on sles-2, the cluster would set
OCF_RESKEY_CRM_meta_notify_start_resource="clone:0 clone:2 clone:3"
OCF_RESKEY_CRM_meta_notify_start_uname="sles-1 sles-3 sles-2"
Example notification variables

10.2.3.7. Proper Interpretation of Notification Environment Variables

Pre-notification (stop)
  • Active resources: $OCF_RESKEY_CRM_meta_notify_active_resource
  • Inactive resources: $OCF_RESKEY_CRM_meta_notify_inactive_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
Post-notification (stop) / Pre-notification (start)
  • Active resources:
$OCF_RESKEY_CRM_meta_notify_active_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Inactive resources:
$OCF_RESKEY_CRM_meta_notify_inactive_resource
plus $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources that were started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources that were stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
Post-notification (start)
  • Active resources:
$OCF_RESKEY_CRM_meta_notify_active_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
plus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Inactive resources:
$OCF_RESKEY_CRM_meta_notify_inactive_resource
plus $OCF_RESKEY_CRM_meta_notify_stop_resource
minus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources that were started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources that were stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource

10.3. Multi-state - Resources That Have Multiple Modes

Multi-state resources are a specialization of Clones (please ensure you understand the section on clones before continuing) that allow the instances to be in one of two operating modes. These modes are called Master and Slave but can mean whatever you wish them to mean. The only limitation is that when an instance is started, it must come up in the Slave state.

10.3.1. Properties

Table 10.5. Properties of a Multi-State Resource
Field Description
id Your name for the multi-state resource

10.3.2. Options

Options inherited from simple resources: priority, target-role, is-managed
Options inherited from clone resources: clone-max, clone-node-max, notify, globally-unique, ordered, interleave
Table 10.6. Multi-state specific resource configuration options
Field Description
master-max How many copies of the resource can be promoted to master status. Defaults to 1.
master-node-max How many copies of the resource can be promoted to master status on a single node. Defaults to 1.

10.3.3. Using Multi-state Resources

10.3.3.1. Instance Attributes

Multi-state resources have no instance attributes, however any that are set here will be inherited by the master's children.

10.3.3.2. Contents

Masters must contain exactly one group or one regular resource.

Warning

You should never reference the name of a master's child. If you think you need to do this, you probably need to re-evaluate your design.

10.3.3.3. Monitoring Multi-State Resources

The normal type of monitor actions you define are not sufficient to monitor a multi-state resource in the Master state. To detect failures of the master instance, you need to define an additional monitor action with role="Master".

Important

It is crucial that every monitor operation has a different interval
Example 10.6. Monitoring both states of a multi-state resource

  <master id="myMasterRsc">
   <primitive id="myRsc" class="ocf" type="myApp" provider="myCorp">
    <operations>
     <op id="public-ip-slave-check" name="monitor" interval="60"/>
     <op id="public-ip-master-check" name="monitor" interval="61" role="Master"/>
    </operations>
   </primitive>
  </master>


10.3.3.4. Constraints

In most cases, a multi-state resources will have a single copy on each active cluster node. However if this is not the case, you can indicate which nodes the cluster should to preferentially assign copies to with resource location constraints. These constraints are written no differently to those for regular resources except that the master's id is used.
When considering multi-state resources in constraints, for most purposes it is sufficient to treat them as clones. The exception is when the rsc-role and/or with-rsc-role (for colocation constraints) and first-action and/or then-action (for ordering constraints) are used.
Table 10.7. Additional constraint options relevant to multi-state resources
Field Description
rsc-role
An additional attribute of colocation constraints that specifies the role that rsc must be in.
Allowed values: Started, Master, Slave
with-rsc-role
An additional attribute of colocation constraints that specifies the role that with-rsc must be in.
Allowed values: Started, Master, Slave
first-action
An additional attribute of ordering constraints that specifies the action that the first resource must complete before executing the specified action for the then resource.
Allowed values: start, stop, promote, demote
then-action
An additional attribute of ordering constraints that specifies the action that the then resource can only execute after the first-action on the first resource has completed.
Allowed values: start, stop, promote, demote. Defaults to the value (specified or implied) of first-action

In the example below, myApp will wait until one of database copies has been started and promoted to master before being started itself. Only if no copies can be promoted will apache-stats be prevented from being active. Additionally, the database will wait for myApp to be stopped before it is demoted.
Colocation of a regular (or group) resource with a multi-state resource means that it can run on any machine with an active copy of the clone that is in the specified state (Master or Slave). In the example, the cluster will choose a location based on where database is running as a Master, and if there are multiple Master instances it will also factor in myApp‘s own location preferences when deciding which location to choose.
Colocation with regular clones and other multi-state resources is also possible. In such cases, the set of allowed locations for the rsc clone is (after role filtering) limited to nodes on which the with-rsc clone is (or will be) in the specified role. Allocation is then performed as-per-normal.
Example 10.7. Example constraints involving multi-state resources

  <constraints>
   <rsc_location id="db-prefers-node1" rsc="database" node="node1" score="500"/>
   <rsc_colocation id="backup-with-db-slave" rsc="backup" with-rsc="database" with-rsc-role="Slave"/>
   <rsc_colocation id="myapp-with-db-master" rsc="myApp" with-rsc="database" with-rsc-role="Master"/>
   <rsc_order id="start-db-before-backup" first="database" then="backup"/>
   <rsc_order id="promote-db-then-app" first="database" first-action="promote" then="myApp" then-action="start"/>
  </constraints>


10.3.3.5. Stickiness

To achieve a stable allocation pattern, clones are slightly sticky by default. If no value for resource-stickiness is provided, the clone will use a value of 1. Being a small value, it causes minimal disturbance to the score calculations of other resources but is enough to prevent Pacemaker from needlessly moving copies around the cluster.

10.3.3.6. Which Resource Instance is Promoted

During the start operation, most Resource Agent scripts should call the crm_master utility. This tool automatically detects both the resource and host and should be used to set a preference for being promoted. Based on this, master-max, and master-node-max, the instance(s) with the highest preference will be promoted.
The other alternative is to create a location constraint that indicates which nodes are most preferred as masters.
Example 10.8. Manually specifying which node should be promoted

  <rsc_location id="master-location" rsc="myMasterRsc">
    <rule id="master-rule" score="100" role="Master">
      <expression id="master-exp" attribute="#uname" operation="eq" value="node1"/>
    </rule>
  </rsc_location>


10.3.3.7. Resource Agent Requirements

Since multi-state resources are an extension of cloned resources, all the requirements of Clones are also requirements of multi-state resources. Additionally, multi-state resources require two extra actions demote and promote. These actions are responsible for changing the state of the resource. Like start and stop, they should return OCF_SUCCESS if they completed successfully or a relevant error code if they did not.
The states can mean whatever you wish, but when the resource is started, it must come up in the mode called Slave. From there the cluster will then decide which instances to promote into a Master.
In addition to the Clone requirements for monitor actions, agents must also accurately report which state they are in. The cluster relies on the agent to report its status (including role) accurately and does not indicate to the agent what role it currently believes it to be in.
Table 10.8. Role implications of OCF return codes
Monitor Return Code Description
OCF_NOT_RUNNING Stopped
OCF_SUCCESS Running (Slave)
OCF_RUNNING_MASTER Running (Master)
OCF_FAILED_MASTER Failed (Master)
Other Failed (Slave)

10.3.3.8. Notifications

Like with clones, supporting notifications requires the notify action to be implemented. Once supported, the notify action will be passed a number of extra variables which, when combined with additional context, can be used to calculate the current state of the cluster and what is about to happen to it.
Table 10.9. Environment variables supplied with Master notify actions [11]
Variable Description
OCF_RESKEY_CRM_meta_notify_type Allowed values: pre, post
OCF_RESKEY_CRM_meta_notify_operation Allowed values: start, stop
OCF_RESKEY_CRM_meta_notify_active_resource Resources the that are running
OCF_RESKEY_CRM_meta_notify_inactive_resource Resources the that are not running
OCF_RESKEY_CRM_meta_notify_master_resource Resources that are running in Master mode
OCF_RESKEY_CRM_meta_notify_slave_resource Resources that are running in Slave mode
OCF_RESKEY_CRM_meta_notify_start_resource Resources to be started
OCF_RESKEY_CRM_meta_notify_stop_resource Resources to be stopped
OCF_RESKEY_CRM_meta_notify_promote_resource Resources to be promoted
OCF_RESKEY_CRM_meta_notify_demote_resource Resources to be demoted
OCF_RESKEY_CRM_meta_notify_start_uname Nodes on which resources will be started
OCF_RESKEY_CRM_meta_notify_stop_uname Nodes on which resources will be stopped
OCF_RESKEY_CRM_meta_notify_promote_uname Nodes on which resources will be promoted
OCF_RESKEY_CRM_meta_notify_demote_uname Nodes on which resources will be demoted
OCF_RESKEY_CRM_meta_notify_active_uname Nodes on which resources are running
OCF_RESKEY_CRM_meta_notify_inactive_uname Nodes on which resources are not running
OCF_RESKEY_CRM_meta_notify_master_uname Nodes on which resources are running in Master mode
OCF_RESKEY_CRM_meta_notify_slave_uname Nodes on which resources are running in Slave mode

10.3.3.9. Proper Interpretation of Notification Environment Variables

Pre-notification (demote)
  • Active resources: $OCF_RESKEY_CRM_meta_notify_active_resource
  • Master resources: $OCF_RESKEY_CRM_meta_notify_master_resource
  • Slave resources: $OCF_RESKEY_CRM_meta_notify_slave_resource
  • Inactive resources: $OCF_RESKEY_CRM_meta_notify_inactive_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources to be demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
Post-notification (demote) / Pre-notification (stop)
  • Active resources: $OCF_RESKEY_CRM_meta_notify_active_resource
  • Master resources:
$OCF_RESKEY_CRM_meta_notify_master_resource
minus $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Slave resources: $OCF_RESKEY_CRM_meta_notify_slave_resource
  • Inactive resources: $OCF_RESKEY_CRM_meta_notify_inactive_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources to be demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources that were demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
Post-notification (stop) / Pre-notification (start)
  • Active resources:
$OCF_RESKEY_CRM_meta_notify_active_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Master resources:
$OCF_RESKEY_CRM_meta_notify_master_resource
minus $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Slave resources:
$OCF_RESKEY_CRM_meta_notify_slave_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Inactive resources:
$OCF_RESKEY_CRM_meta_notify_inactive_resource
plus $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources to be demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources that were demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources that were stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
Post-notification (start) / Pre-notification (promote)
  • Active resources:
$OCF_RESKEY_CRM_meta_notify_active_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
plus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Master resources:
$OCF_RESKEY_CRM_meta_notify_master_resource
minus $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Slave resources:
$OCF_RESKEY_CRM_meta_notify_slave_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
plus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Inactive resources:
$OCF_RESKEY_CRM_meta_notify_inactive_resource
plus $OCF_RESKEY_CRM_meta_notify_stop_resource
minus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources to be demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources that were started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources that were demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources that were stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
Post-notification (promote)
  • Active resources:
$OCF_RESKEY_CRM_meta_notify_active_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
plus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Master resources:
$OCF_RESKEY_CRM_meta_notify_master_resource
minus $OCF_RESKEY_CRM_meta_notify_demote_resource
plus $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Slave resources:
$OCF_RESKEY_CRM_meta_notify_slave_resource
minus $OCF_RESKEY_CRM_meta_notify_stop_resource
plus $OCF_RESKEY_CRM_meta_notify_start_resource
minus $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Inactive resources:
$OCF_RESKEY_CRM_meta_notify_inactive_resource
plus $OCF_RESKEY_CRM_meta_notify_stop_resource
minus $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources to be promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources to be demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources to be stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource
  • Resources that were started: $OCF_RESKEY_CRM_meta_notify_start_resource
  • Resources that were promoted: $OCF_RESKEY_CRM_meta_notify_promote_resource
  • Resources that were demoted: $OCF_RESKEY_CRM_meta_notify_demote_resource
  • Resources that were stopped: $OCF_RESKEY_CRM_meta_notify_stop_resource


[11] Variables in bold are specific to Master resources and all behave in the same manner as described for Clone resources.

Chapter 11. Protecting Your Data - STONITH

11.1. Why You Need STONITH

STONITH is an acronym for Shoot-The-Other-Node-In-The-Head and it protects your data from being corrupted by rouge nodes or concurrent access.
Just because a node is unresponsive, this doesn't mean it isn't accessing your data. The only way to be 100% sure that your data is safe, is to use STONITH so we can be certain that the node is truly offline, before allowing the data to be accessed from another node.
STONITH also has a role to play in the event that a clustered service cannot be stopped. In this case, the cluster uses STONITH to force the whole node offline, thereby making it safe to start the service elsewhere.

11.2. What STONITH Device Should You Use

It is crucial that the STONITH device can allow the cluster to differentiate between a node failure and a network one.
The biggest mistake people make in choosing a STONITH device is to use remote power switch (such as many on-board IMPI controllers) that shares power with the node it controls. In such cases, the cluster cannot be sure if the node is really offline, or active and suffering from a network fault.
Likewise, any device that relies on the machine being active (such as SSH-based "devices" used during testing) are inappropriate.

11.3. Configuring STONITH

  1. Find the correct driver: stonith -L
  2. Since every device is different, the parameters needed to configure it will vary. To find out the parameters required by the device: stonith -t type -n
    Hopefully the developers chose names that make sense, if not you can query for some additional information by finding an active cluster node and running:
    lrmadmin -M stonith type pacemaker
    The output should be XML formatted text containing additional parameter descriptions
  3. Create a file called stonith.xml containing a primitive resource with a class of stonith, a type of type and a parameter for each of the values returned in step 2
  4. Create a clone from the primitive resource if the device can shoot more than one node and supports multiple simultaneous connections.
  5. Upload it into the CIB using cibadmin: cibadmin -C -o resources --xml-file stonith.xml

11.3.1. Example

Assuming we have an IBM BladeCenter consisting of four nodes and the management interface is active on 10.0.0.1, then we would chose the external/ibmrsa driver in step 2 and obtain the following list of parameters
	     stonith -t external/ibmrsa -n 
	     hostname ipaddr userid passwd type 
Figure 11.1. Obtaining a list of STONITH Parameters

from which we would create a STONITH resource fragment that might look like this
Example 11.1. Sample STONITH Resource

      <clone id="Fencing">
       <meta_attributes id="fencing">
         <nvpair id="Fencing-unique" name="globally-unique" value="false"/>
       </meta_attributes>
       <primitive id="rsa" class="stonith" type="external/ibmrsa">
        <operations>
         <op id="rsa-mon-1" name="monitor" interval="120s"/>
        </operations>
        <instance_attributes id="rsa-parameters">
          <nvpair id="rsa-attr-1" name="hostname" value="node1 node2 node3 node4"/>
          <nvpair id="rsa-attr-1" name="ipaddr" value="10.0.0.1"/>
          <nvpair id="rsa-attr-1" name="userid" value="testuser"/>
          <nvpair id="rsa-attr-1" name="passwd" value="abc123"/>
          <nvpair id="rsa-attr-1" name="type" value="ibm"/>
        </instance_attributes>
       </primitive>
      </clone>


Chapter 12. Status - Here be dragons

Most users never need understand the contents of the status section and can be content with the output from crm_mon. However for those with a curious inclination, the following attempts to proved an overview of its contents.

12.1. Node Status

In addition to the cluster's configuration, the CIB holds an up-to-date representation of each cluster node in the status section.

  <node_state id="cl-virt-1" uname="cl-virt-2" ha="active" in_ccm="true" crmd="online" join="member" expected="member" crm-debug-origin="do_update_resource">
   <transient_attributes id="cl-virt-1"/>
   <lrm id="cl-virt-1"/>
  </node_state>

Figure 12.1. A bare-bones status entry for a healthy node called cl-virt-1

Users are highly recommended not to modify any part of a node's state directly. The cluster will periodically regenerate the entire section from authoritative sources. So any changes should be with the tools for those subsystems.
Table 12.1. Authoritative Sources for State Information
Dataset Authoritative Source
node_state fields crmd
transient_attributes tag attrd
lrm tag lrmd

The fields used in the node_state objects are named as they are largely for historical reasons and are rooted in Pacemaker's origins as the Heartbeat resource manager. They have remained unchanged to preserve compatibility with older versions.
Table 12.2. Node Status Fields
Field Description
id Unique identifier for the node. Corosync based clusters use the same value as uname, Heartbeat cluster use a human-readable (but annoying) UUID.
uname The node's machine name (output from uname -n)
ha Is the cluster software active on the node. Allowed values: active, dead
in_ccm Is the node part of the cluster's membership. Allowed values: true, false
crmd Is the crmd process active on the node. Allowed values: online, offline
join Is the node participating in hosting resources. Allowed values: down, pending, member, banned
expected Expected value for join
crm-debug-origin Diagnostic indicator. The origin of the most recent change(s).

The cluster uses these fields to determine if, at the node level, the node is healthy or is in a failed state and needs to be fenced.

12.2. Transient Node Attributes

Like regular node attributes, the name/value pairs listed here also help describe the node. However they are forgotten by the cluster when the node goes offline. This can be useful, for instance, when you only want a node to be in standby mode (not able to run resources) until the next reboot.
In addition to any values the administrator sets, the cluster will also store information about failed resources here.

     <transient_attributes id="cl-virt-1">
      <instance_attributes id="status-cl-virt-1">
       <nvpair id="status-cl-virt-1-pingd" name="pingd" value="3"/>
       <nvpair id="status-cl-virt-1-probe_complete" name="probe_complete" value="true"/>
       <nvpair id="status-cl-virt-1-fail-count-pingd:0" name="fail-count-pingd:0" value="1"/>
       <nvpair id="status-cl-virt-1-last-failure-pingd:0" name="last-failure-pingd:0" value="1239009742"/>
      </instance_attributes>
     </transient_attributes>

Figure 12.2. Example set of transient node attributes for node "cl-virt-1"

In the above example, we can see that the pingd:0 resource has failed once, at Mon Apr 6 11:22:22 2009. [12] We also see that the node is connected to three "pingd" peers and that all known resources have been checked for on this machine (probe_complete).

12.3. Operation History

A node's resource history is held in the lrm_resources tag (a child of the lrm tag). The information stored here includes enough information for the cluster to stop the resource safely if it is removed from the configuration section. Specifically we store the resource's id, class, type and provider.
  <lrm_resource id="apcstonith" type="apcmastersnmp" class="stonith">
Figure 12.3. A record of the apcstonith resource

Additionally, we store the last job for every combination of resource, action and interval. The concatenation of the values in this tuple are used to create the id of the lrm_rsc_op object.
Table 12.3. Contents of an lrm_rsc_op job.
Field Description
id Identifier for the job constructed from the resource id, operation and interval.
call-id The job's ticket number. Used as a sort key to determine the order in which the jobs were executed.
operation The action the resource agent was invoked with.
interval The frequency, in milliseconds, at which the operation will be repeated. 0 indicates a one-off job.
op-status The job's status. Generally this will be either 0 (done) or -1 (pending). Rarely used in favor of rc-code.
rc-code The job's result. Refer to Section B.3, “How Does the Cluster Interpret the OCF Return Codes?” for details on what the values here mean and how they are interpreted.
last-run Diagnostic indicator. Machine local date/time, in seconds since epoch, at which the job was executed.
last-rc-change Diagnostic indicator. Machine local date/time, in seconds since epoch, at which the job first returned the current value of rc-code
exec-time Diagnostic indicator. Time, in seconds, that the job was running for
queue-time Diagnostic indicator. Time, in seconds, that the job was queued for in the LRMd
crm_feature_set The version which this job description conforms to. Used when processing op-digest
transition-key A concatenation of the job's graph action number, the graph number, the expected result and the UUID of the crmd instance that scheduled it. This is used to construct transition-magic (below).
transition-magic A concatenation of the job's op-status, rc-code and transition-key. Guaranteed to be unique for the life of the cluster (which ensures it is part of CIB update notifications) and contains all the information needed for the crmd to correctly analyze and process the completed job. Most importantly, the decomposed elements tell the crmd if the job entry was expected and whether it failed.
op-digest An MD5 sum representing the parameters passed to the job. Used to detect changes to the configuration and restart resources if necessary.
crm-debug-origin Diagnostic indicator. The origin of the current values.

12.3.1. Simple Example


  <lrm_resource id="apcstonith" type="apcmastersnmp" class="stonith"> 
    <lrm_rsc_op id="apcstonith_monitor_0" operation="monitor" call-id="2" rc-code="7" op-status="0" interval="0" 
		crm-debug-origin="do_update_resource" crm_feature_set="3.0.1" 
		op-digest="2e3da9274d3550dc6526fb24bfcbcba0"
		transition-key="22:2:7:2668bbeb-06d5-40f9-936d-24cb7f87006a"
		transition-magic="0:7;22:2:7:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		last-run="1239008085" last-rc-change="1239008085" exec-time="10" queue-time="0"/>
  </lrm_resource>

Figure 12.4. A monitor operation performed by the cluster to determine the current state of the apcstonith resource

In the above example, the job is a non-recurring monitor often referred to as a "probe" for the apcstonith resource. The cluster schedules probes for every configured resource on when a new node starts, in order to determine the resource's current state before it takes further any further action.
From the transition-key, we can see that this was the 22nd action of the 2nd graph produced by this instance of the crmd (2668bbeb-06d5-40f9-936d-24cb7f87006a). The third field of the transition-key contains a 7, this indicates that the job expects to find the resource inactive. By now looking at the rc-code property, we see that this was the case.
Evidently, the cluster started the resource elsewhere as that is the only job recorded for this node.

12.3.2. Complex Resource History Example


  <lrm_resource id="pingd:0" type="pingd" class="ocf" provider="pacemaker">
    <lrm_rsc_op id="pingd:0_monitor_30000" operation="monitor" call-id="34" rc-code="0" op-status="0" interval="30000" 
		crm-debug-origin="do_update_resource" crm_feature_set="3.0.1" 
		op-digest="a0f8398dac7ced82320fe99fd20fbd2f"
		transition-key="10:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		transition-magic="0:0;10:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		last-run="1239009741" last-rc-change="1239009741" exec-time="10" queue-time="0"/>
    <lrm_rsc_op id="pingd:0_stop_0" operation="stop" 
		crm-debug-origin="do_update_resource" crm_feature_set="3.0.1" call-id="32" rc-code="0" op-status="0" interval="0" 
		op-digest="313aee7c6aad26e290b9084427bbab60"
		transition-key="11:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		transition-magic="0:0;11:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		last-run="1239009741" last-rc-change="1239009741" exec-time="10" queue-time="0"/>
    <lrm_rsc_op id="pingd:0_start_0" operation="start" call-id="33" rc-code="0" op-status="0" interval="0" 
		crm-debug-origin="do_update_resource" crm_feature_set="3.0.1" 
		op-digest="313aee7c6aad26e290b9084427bbab60"
		transition-key="31:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		transition-magic="0:0;31:11:0:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		last-run="1239009741" last-rc-change="1239009741" exec-time="10" queue-time="0" />
    <lrm_rsc_op id="pingd:0_monitor_0" operation="monitor" call-id="3" rc-code="0" op-status="0" interval="0" 
		crm-debug-origin="do_update_resource" crm_feature_set="3.0.1" 
		op-digest="313aee7c6aad26e290b9084427bbab60"
		transition-key="23:2:7:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		transition-magic="0:0;23:2:7:2668bbeb-06d5-40f9-936d-24cb7f87006a" 
		last-run="1239008085" last-rc-change="1239008085" exec-time="20" queue-time="0"/>
  </lrm_resource>

Figure 12.5. Resource history of a pingd clone with multiple jobs

When more than one job record exists, it is important to first sort them by call-id before interpret them. Once sorted, the above example can be summarized as:
  1. A non-recurring monitor operation returning 7 (not running), with a call-id of 3
  2. A stop operation returning 0 (success), with a call-id of 32
  3. A start operation returning 0 (success), with a call-id of 33
  4. A recurring monitor returning 0 (success), with a call-id of 34
The cluster processes each job record to build up a picture of the resource's state. After the first and second entries, it is considered stopped and after the third it considered active. Based on the last operation, we can tell that the resource is currently active.
Additionally, from the presence of a stop operation with a lower call-id than that of the start operation, we can conclude that the resource has been restarted. Specifically this occurred as part of actions 11 and 31 of transition 11 from the crmd instance with the key 2668bbeb-06d5-40f9-936d-24cb7f87006a. This information can be helpful for locating the relevant section of the logs when looking for the source of a failure.


[12] You can use the following Perl one-liner to print a human readable of any seconds-since-epoch value:

perl -e 'print scalar(localtime($seconds))."\n"'

FAQ

A.1. History

Why is the Project Called Pacemaker?
First of all, the reason its not called the CRM is because of the abundance of terms that are commonly abbreviated to those three letters.
The Pacemaker name came from Kham, a good friend of mine, and was originally used by a Java GUI that I was prototyping in early 2007. Alas other commitments have prevented the GUI from progressing much and, when it came time to choose a name for this project, Lars suggested it was an even better fit for an independent CRM.
The idea stems from the analogy between the role of this software and that of the little device that keeps the human heart pumping. Pacemaker monitors the cluster and intervenes when necessary to ensure the smooth operation of the services it provides.
There were a number of other names (and acronyms) tossed around, but suffice to say "Pacemaker" was the best
Why was the Pacemaker Project Created?
The decision was made to spin-off the CRM into its own project after the 2.1.3 Heartbeat release in order to
  • support both the Corosync and Heartbeat cluster stacks equally
  • decouple the release cycles of two projects at very different stages of their life-cycles
  • foster the clearer package boundaries, thus leading to
  • better and more stable interfaces

A.2. Setup

What Messaging Layers are Supported?
Can I Choose which Messaging Layer to use at Run Time?
Yes. The CRM will automatically detect who started it and behave accordingly.
Can I Have a Mixed Heartbeat-Corosync Cluster?
No.
Which Messaging Layer Should I Choose?
This is discussed in Appendix D, Installation.
Where Can I Get Pre-built Packages?
Official packages for most major .rpm and based distributions are available from:
For Debian packages, building from source and details on using the above repositories, see our installation page.
What Versions of Pacemaker Are Supported?
Please refer to the Releases page for an up-to-date list of versions supported directly by the project.
When seeking assistance, please try to ensure you have one of these versions.

More About OCF Resource Agents

B.1. Location of Custom Scripts

OCF Resource Agents are found in /usr/lib/ocf/resource.d/provider.
When creating your own agents, you are encouraged to create a new directory under /usr/lib/ocf/resource.d/ so that they are not confused with (or overwritten by) the agents shipped with Heartbeat. So, for example, if you chose the provider name of bigCorp and wanted a new resource named bigApp, you would create a script called /usr/lib/ocf/resource.d/bigCorp/bigApp and define a resource:
<primitive id="custom-app" class="ocf" provider="bigCorp" type="bigApp"/>

B.2. Actions

All OCF Resource Agents are required to implement the following actions
Table B.1. Required Actions for OCF Agents
Action Description Instructions
start Start the resource Return 0 on success and an appropriate error code otherwise. Must not report success until the resource is fully active.
stop Stop the resource Return 0 on success and an appropriate error code otherwise. Must not report success until the resource is fully stopped.
monitor Check the resource's state
Exit 0 if the resource is running, 7 if it is stopped and anything else if it is failed.
NOTE: The monitor script should test the state of the resource on the local machine only.
meta-data Describe the resource
Provide information about this resource as an XML snippet. Exit with 0.
NOTE: This is not performed as root.
validate-all Verify the supplied parameters are correct Exit with 0 if parameters are valid, 2 if not valid, 6 if resource is not configured.

Additional requirements (not part of the OCF specs) are placed on agents that will be used for advanced concepts like clones and multi-state resources.
Table B.2. Optional Actions for OCF Agents
Action Description Instructions
promote Promote the local instance of a multi-state resource to the master/primary state Return 0 on success
demote Demote the local instance of a multi-state resource to the slave/secondary state Return 0 on success
notify Used by the cluster to send the agent pre and post notification events telling the resource what is or did just take place Must not fail. Must exit 0

Some actions specified in the OCF specs are not currently used by the cluster
  • reload - reload the configuration of the resource instance without disrupting the service
  • recover - a variant of the start action, this should try to recover a resource locally.
Remember to use ocf-tester to verify that your new agent complies with the OCF standard properly.

B.3. How Does the Cluster Interpret the OCF Return Codes?

The first thing the cluster does is check the return code against the expected result. If the result does not match the expected value, then the operation is considered to have failed and recovery action is initiated.
There are three types of failure recovery:
Table B.3. Types of recovery performed by the cluster
Recovery Type Description Action Taken by the Cluster
soft A transient error occurred Restart the resource or move it to a new location
hard A non-transient error that may be specific to the current node occurred Move the resource elsewhere and prevent it from being retried on the current node
fatal A non-transient error that will be common to all cluster nodes (I.e. a bad configuration was specified) Stop the resource and prevent it from being started on any cluster node

Assuming an action is considered to have failed, the following table outlines the different OCF return codes and the type of recovery the cluster will initiate when it is received.
Table B.4. OCF Return Codes and How They are Handled
OCF Return Code OCF Alias Description Recovery Type
0 OCF_SUCCESS Success. The command complete successfully. This is the expected result for all start, stop, promote and demote commands. soft
1 OCF_ERR_GENERIC Generic "there was a problem" error code. soft
2 OCF_ERR_ARGS The resource's configuration is not valid on this machine. Eg. Refers to a location/tool not found on the node. hard
3 OCF_ERR_UNIMPLEMENTED The requested action is not implemented. hard
4 OCF_ERR_PERM The resource agent does not have sufficient privileges to complete the task. hard
5 OCF_ERR_INSTALLED The tools required by the resource are not installed on this machine. hard
6 OCF_ERR_CONFIGURED The resource's configuration is invalid. Eg. A required parameters are missing. fatal
7 OCF_NOT_RUNNING The resource is safely stopped. The cluster will not attempt to stop a resource that returns this for any action. N/A
8 OCF_RUNNING_MASTER The resource is running in Master mode. soft
9 OCF_FAILED_MASTER The resource is in Master mode but has failed. The resource will be demoted, stopped and then started (and possibly promoted) again. soft
other NA Custom error code. soft

Although counter intuitive, even actions that return 0 (aka. OCF_SUCCESS) can be considered to have failed. This can happen when a resource that is expected to be in the Master state is found running as a Slave, or when a resource is found active on multiple machines..

B.3.1. Exceptions

  • Non-recurring monitor actions (probes) that find a resource active (or in Master mode) will not result in recovery action unless it is also found active elsewhere
  • The recovery action taken when a resource is found active more than once is determined by the multiple-active property of the resource
  • Recurring actions that return OCF_ERR_UNIMPLEMENTED do not cause any type of recovery

What Changed in 1.0

C.1. New

C.2. Changed

  • Syntax
    • All resource and cluster options now use dashes (-) instead of underscores (_)
    • master_slave was renamed to master
    • The attributes container tag was removed
    • The operation field pre-req has been renamed requires
    • All operations must have an interval, start/stop must have it set to zero
  • The stonith-enabled option now defaults to true.
  • The cluster will refuse to start resources if stonith-enabled is true (or unset) and no STONITH resources have been defined
  • The attributes of colocation and ordering constraints were renamed for clarity. See Section 6.3, “Specifying the Order Resources Should Start/Stop In” and Section 6.4, “Placing Resources Relative to other Resources”
  • resource-failure-stickiness has been replaced by migration-threshold. See Section 9.3.2, “Moving Resources Due to Failure”
  • The arguments for command-line tools has been made consistent
  • Switched to RelaxNG schema validation and libxml2 parser.
    • id fields are now XML IDs which have the following limitations
      • id's cannot contain colons (:)
      • id's cannot begin with a number
      • id's must be globally unique (not just unique for that tag)
    • Some fields (such as those in constraints that refer to resources) are IDREFs. This means that they must reference existing resources or objects in order for the configuration to be valid. Removing an object which is referenced elsewhere will therefor fail.
    • The CIB representation from which the MD5 digest used to verify CIBs has changed. This means that every CIB update will require a full refresh on any upgraded nodes until the cluster is fully upgraded to 1.0. This will result in significant performance degradation and it is therefor highly inadvisable to run a mixed 1.0/0.6 cluster for any longer than absolutely necessary.
  • Ping node information no longer needs to be added to ha.cf Simply include the lists of hosts in your ping resource(s).

C.3. Removed

Installation

D.1. Choosing a Cluster Stack

Ultimately the choice of cluster stack is a personal decision that must be made in the context of you or your company's needs and strategic direction. Pacemaker currently functions equally well with both stacks.
Here are some factors that may influence the decision
  • SUSE/Novell, Red Hat and Oracle are all putting their collective weight behind the Corosync cluster stack.
  • Corosync is an OSI Certified implementation of an industry standard (the Service Availability Forum Application Interface Specification).
  • Using Corosync gives your applications access to the following additional cluster services
    • checkpoint service
    • distributed locking service
    • extended virtual synchrony service
    • cluster closed process group service
  • It is likely that Pacemaker, at some point in the future, will make use of some of these additional services not provided by Heartbeat
  • To date, Pacemaker has received less real-world testing on Corosync than it has on Heartbeat.

D.2. Enabling Pacemaker

D.2.1. For Corosync

The Corosync configuration is normally located in /etc/corosync/corosync.conf and an example for a machine with an address of 1.2.3.4 in a cluster communicating on port 1234 (without peer authentication and message encryption) is shown below.
Example D.1. An example Corosync configuration file
  totem {
      version: 2
      secauth: off
      threads: 0
      interface {
          ringnumber: 0
          bindnetaddr: 1.2.3.4
          mcastaddr: 226.94.1.1
          mcastport: 1234
      }
  }
  logging {
      fileline: off
      to_syslog: yes
      syslog_facility: daemon
  }
  amf {
      mode: disabled
  }

The logging should be mostly obvious and the amf section refers to the Availability Management Framework and is not covered in this document.
The interesting part of the configuration is the totem section. This is where we define the how the node can communicate with the rest of the cluster and what protocol version and options (including encryption[13]) it should use. Beginners are encouraged to use the values shown and modify the interface section based on their network.
It is also possible to configure Corosync for an IPv6 based environment. Simply configure bindnetaddr and mcastaddr with their IPv6 equivalents. Eg
Example D.2. Example options for an IPv6 environment
  bindnetaddr: fec0::1:a800:4ff:fe00:20 
  mcastaddr: ff05::1

To tell Corosync to use the Pacemaker cluster manager, add the following fragment to a functional Corosync configuration and restart the cluster.
Example D.3. Configuration fragment for enabling Pacemaker under Corosync
  aisexec {
    user:  root
    group: root
  }
  service {
    name: pacemaker
    ver: 0
  }

The cluster needs to be run as root so that its child processes (the lrmd in particular) have sufficient privileges to perform the actions requested of it. After-all, a cluster manager that can't add an IP address or start apache is of little use.
The second directive is the one that actually instructs the cluster to run Pacemaker.

D.2.2. For Heartbeat

Add the following to a functional ha.cf configuration file and restart Heartbeat
Example D.4. Configuration fragment for enabling Pacemaker under Heartbeat
  crm respawn



[13] Please consult the Corosync website and documentation for details on enabling encryption and peer authentication for the cluster.

Upgrading Cluster Software

E.1. Version Compatibility

When releasing newer versions we take care to make sure we are backwardly compatible with older versions. While you will always be able to upgrade from version x to x+1, in order to continue to produce high quality software it may occasionally be necessary to drop compatibility with older versions.
There will always be an upgrade path from any series-2 release to any other series-2 release.
There are three approaches to upgrading your cluster software
  • Complete Cluster Shutdown
  • Rolling (node by node)
  • Disconnect and Reattach
Each method has advantages and disadvantages, some of which are listed in the table below, and you should chose the one most appropriate to your needs.
Table E.1. Summary of Upgrade Methodologies
Type Available between all software versions Service Outage During Upgrade Service Recovery During Upgrade Exercises Failover Logic/Configuration Allows change of cluster stack type [a]
Shutdown yes always N/A no yes
Rolling no always yes yes no
Reattach yes only due to failure no no yes

[a] For example, switching from Heartbeat to Corosync. Consult the Heartbeat or Corosync documentation to see if upgrading them to a newer version is also supported


E.2. Complete Cluster Shutdown

In this scenario one shuts down all cluster nodes and resources and upgrades all the nodes before restarting the cluster.

E.2.1. Procedure

  1. On each node:
    1. Shutdown the cluster stack (Heartbeat or Corosync)
    2. Upgrade the Pacemaker software. This may also include upgrading the cluster stack and/or the underlying operating system..
  2. Check the configuration manually or with the crm_verify tool if available.
  3. On each node:
    1. Start the cluster stack. This can be either Corosync or Heartbeat and does not need to be the same as the previous cluster stack.

E.3. Rolling (node by node)

In this scenario each node is removed from the cluster, upgraded and then brought back online until all nodes are running the newest version.

Important

This method is currently broken between Pacemaker 0.6.x and 1.0.x
Measures have been put into place to ensure rolling upgrades always work for versions after 1.0.0 If there is sufficient demand, the work to repair 0.6 -> 1.0 compatibility will be carried out. Otherwise, please try one of the other upgrade strategies. Detach/Reattach is a particularly good option for most people.

E.3.1. Procedure

On each node:
  1. Shutdown the cluster stack (Heartbeat or Corosync)
  2. Upgrade the Pacemaker software. This may also include upgrading the cluster stack and/or the underlying operating system.
    1. On the first node, check the configuration manually or with the crm_verify tool if available.
  3. Start the cluster stack. This must be the same type of cluster stack (Corosync or Heartbeat) that the rest of the cluster is using. Upgrading Corosync/Heartbeat may also be possible, please consult the documentation for those projects to see if the two versions will be compatible.
Repeat for each node in the cluster

E.3.2. Version Compatibility

Table E.2. Version Compatibility Table
Version being Installed Oldest Compatible Version
Pacemaker 1.0.x Pacemaker 1.0.0
Pacemaker 0.7.x Pacemaker 0.6 or Heartbeat 2.1.3
Pacemaker 0.6.x Heartbeat 2.0.8
Heartbeat 2.1.3 (or less) Heartbeat 2.0.4
Heartbeat 2.0.4 (or less) Heartbeat 2.0.0
Heartbeat 2.0.0 None. Use an alternate upgrade strategy.

E.3.3. Crossing Compatibility Boundaries

Rolling upgrades that cross compatibility boundaries must be preformed in multiple steps. For example, to perform a rolling update from Heartbeat 2.0.1 to Pacemaker 0.6.6 one must:
  1. Perform a rolling upgrade from Heartbeat 2.0.1 to Heartbeat 2.0.4
  2. Perform a rolling upgrade from Heartbeat 2.0.4 to Heartbeat 2.1.3
  3. Perform a rolling upgrade from Heartbeat 2.1.3 to Pacemaker 0.6.6

E.4. Disconnect and Reattach

A variant of a complete cluster shutdown, but the resources are left active and re-detected when the cluster is restarted.

E.4.1. Procedure

  1. Tell the cluster to stop managing services. This is required to allow the services to remain active after the cluster shuts down.
    crm_attribute -t crm_config -n is-managed-default -v false
  2. For any resource that has a value for is-managed, make sure it is set to false (so that the cluster will not stop it)
    crm_resource -t primitive -r <rsc_id> -p is-managed -v false
  3. On each node:
    1. Shutdown the cluster stack (Heartbeat or Corosync)
    2. Upgrade the cluster stack program - This may also include upgrading the underlying operating system.
  4. Check the configuration manually or with the crm_verify tool if available.
  5. On each node:
    1. Start the cluster stack. This can be either Corosync or Heartbeat and does not need to be the same as the previous cluster stack.
  6. Verify the cluster re-detected all resources correctly
  7. Allow the cluster to resume managing resources again
    crm_attribute -t crm_config -n is-managed-default -v true
  8. For any resource that has a value for is-managed reset it to true (so the cluster can recover the service if it fails) if desired
    crm_resource -t primitive -r <rsc_id> -p is-managed -v false

E.4.2. Notes

Important

Always check your existing configuration is still compatible with the version you are installing before starting the cluster.

Note

The oldest version of the CRM to support this upgrade type was in Heartbeat 2.0.4

Upgrading the Configuration from 0.6

F.1. Preparation

Download the latest DTD from http://hg.clusterlabs.org/pacemaker/stable-1.0/file-raw/tip/xml/crm.dtd and ensure your configuration validates.

F.2. Perform the upgrade

F.2.1. Upgrade the software

F.2.2. Upgrade the Configuration

As XML is not the friendliest of languages, it is common for cluster administrators to have scripted some of their activities. In such cases, it is likely that those scripts will not work with the new 1.0 syntax.
In order to support such environments, it is actually possible to continue using the old 0.6 syntax.
The downside however, is that not all the new features will be available and there is a performance impact since the cluster must do a non-persistent configuration upgrade before each transition. So while using the old syntax is possible, it is not advisable to continue using it indefinitely.
Even if you wish to continue using the old syntax, it is advisable to follow the upgrade procedure to ensure that the cluster is able to use your existing configuration (since it will perform much the same task internally).
  1. Create a shadow copy to work with
    crm_shadow --create upgrade06
  2. Verify the configuration is valid
    crm_verify --live-check
  3. Fix any errors or warnings
  4. Perform the upgrade
    cibadmin --upgrade
    If this step fails, there are three main possibilities
    1. The configuration was not valid to start with - go back to step 2
    2. The transformation failed - report a bug or email the project at pacemaker@oss.clusterlabs.org
    3. The transformation was successful but produced an invalid result [14]
    If the result of the transformation is invalid, you may see a number of errors from the validation library. If these are not helpful, visit http://clusterlabs.org/wiki/Validation_FAQ and/or try the following procedure described below under Section F.2.3, “Manually Upgrading the Configuration”.
  5. Check the changes
    crm_shadow --diff
    If at this point there is anything about the upgrade that you wish to fine-tune (for example, to change some of the automatic IDs) now is the time to do so. Since the shadow configuration is not in use by the cluster, it is safe to edit the file manually:
    crm_shadow --edit
    Will open the configuration in your favorite editor (or whichever one is specified by the standard EDITOR environment variable).
  6. Preview how the cluster will react
    Test what the cluster will do when you upload the new configuration
    ptest -VVVVV --live-check --save-dotfile upgrade06.dot
    graphviz upgrade06.dot
    Verify that either no resource actions will occur or that you are happy with any that are scheduled. If the output contains actions you do not expect (possibly due to changes to the score calculations), you may need to make further manual changes. See Section 2.7, “Testing Your Configuration Changes” for further details on how to interpret the output of ptest.
  7. Upload the changes
    crm_shadow --commit upgrade06 --force
    If this step fails, something really strange has occurred. You should report a bug.

F.2.3. Manually Upgrading the Configuration

It is also possible to perform the configuration upgrade steps manually. To do this
  1. Locate the upgrade06.xsl conversion script or download the latest version from version control
  2. xsltproc /path/tp/upgrade06.xsl config06.xml > config10.xml
  3. Locate the pacemaker.rng script.
  4. xmllint --relaxng /path/tp/pacemaker.rng config10.xml
The advantage of this method is that it can be performed without the cluster running and any validation errors should be more informative (despite being generated by the same library!) since they include line numbers.


[14] The most common reason is ID values being repeated or invalid. Pacemaker 1.0 is much stricter regarding this type of validation

Is This init Script LSB Compatible?

The relevant part of LSB spec can be found at: http://refspecs.freestandards.org/LSB_3.1.0/LSB-Core-generic/LSB-Core-generic/iniscrptact.html. It includes a description of all the return codes listed here.
Assuming some_service is configured correctly and currently not active, the following sequence will help you determine if it is LSB compatible:
  1. Start (stopped):
    /etc/init.d/some_service start ; echo "result: $?"
    1. Did the service start?
    2. Did the command print result: 0 (in addition to the regular output)?
  2. Status (running):
    /etc/init.d/some_service status ; echo "result: $?"
    1. Did the script accept the command?
    2. Did the script indicate the service was running?
    3. Did the command print result: 0 (in addition to the regular output)?
  3. Start (running):
    /etc/init.d/some_service start ; echo "result: $?"
    1. Is the service still running?
    2. Did the command print result: 0 (in addition to the regular output)?
  4. Stop (running):
    /etc/init.d/some_service stop ; echo "result: $?"
    1. Was the service stopped?
    2. Did the command print result: 0 (in addition to the regular output)?
  5. Status (stopped):
    /etc/init.d/some_service status ; echo "result: $?"
    1. Did the script accept the command?
    2. Did the script indicate the service was not running?
    3. Did the command print result: 3 (in addition to the regular output)?
  6. Stop (stopped):
    /etc/init.d/some_service stop ; echo "result: $?"
    1. Is the service still stopped?
    2. Did the command print result: 0 (in addition to the regular output)?
  7. Status (failed):
    This step is not readily testable and relies on manual inspection of the script.
    The script can use one of the error codes (other than 3) listed in the LSB spec to indicate that it is active but failed.
    This tells the cluster that before moving the resource to another node, it needs to stop it on the existing one first.
The script can use one of the error codes (other than 3) listed in the LSB spec to indicate that it is active but failed. This tells the cluster that before moving the resource to another node, it needs to stop it on the existing one first.
If the answer to any of the above questions is no, then the script is not LSB compliant. Your options are then to either fix the script or write an OCF agent based on the existing script.

Sample Configurations

H.1. An Empty Configuration

Example H.1. An empty configuration

  <cib admin_epoch="0" epoch="0" num_updates="0" have-quorum="false">
   <configuration>
    <crm_config/>
    <nodes/>
    <resources/>
    <constraints/>
   </configuration>
   <status/>
  </cib>


H.2. A Simple Configuration

Example H.2. 2 nodes, some cluster options and a resource

  <cib admin_epoch="0" epoch="1" num_updates="0" have-quorum="false" validate-with="pacemaker-1.0">
    <configuration>
      <crm_config>
        <nvpair id="option-1" name="symmetric-cluster" value="true"/>
        <nvpair id="option-2" name="no-quorum-policy" value="stop"/>
      </crm_config>
      <op_defaults>
        <nvpair id="op-default-1" name="timeout" value="30s"/>
      </op_defaults>
      <rsc_defaults>
        <nvpair id="rsc-default-1" name="resource-stickiness" value="100"/>
        <nvpair id="rsc-default-2" name="migration-threshold" value="10"/>
      </rsc_defaults>
      <nodes>
       <node id="xxx" uname="c001n01" type="normal"/>
       <node id="yyy" uname="c001n02" type="normal"/>
      </nodes>
      <resources>
        <primitive id="myAddr" class="ocf" provider="heartbeat" type="IPaddr">
          <operations>
           <op id="myAddr-monitor" name="monitor" interval="300s"/>
          </operations>
          <instance_attributes>
             <nvpair name="ip" value="10.0.200.30"/>
          </instance_attributes>
        </primitive>
      </resources>
      <constraints>
       <rsc_location id="myAddr-prefer" rsc="myAddr" node="c001n01" score="INFINITY"/>
      </constraints>
    </configuration>
    <status/>
  </cib>


In this example, we have one resource (an IP address) that we check every five minutes and will run on host c001n01 until either the resource fails 10 times or the host shuts down.

H.3. An Advanced Configuration

Example H.3. groups and clones with stonith

  <cib admin_epoch="0" epoch="1" num_updates="0" have-quorum="false" validate-with="pacemaker-1.0">
    <configuration>
      <crm_config>
        <nvpair id="option-1" name="symmetric-cluster" value="true"/>
        <nvpair id="option-2" name="no-quorum-policy" value="stop"/>
        <nvpair id="option-3" name="stonith-enabled" value="true"/>
      </crm_config>
      <op_defaults>
        <nvpair id="op-default-1" name="timeout" value="30s"/>
      </op_defaults>
      <rsc_defaults>
        <nvpair id="rsc-default-1" name="resource-stickiness" value="100"/>
        <nvpair id="rsc-default-2" name="migration-threshold" value="10"/>
      </rsc_defaults>
      <nodes>
       <node id="xxx" uname="c001n01" type="normal"/>
       <node id="yyy" uname="c001n02" type="normal"/>
       <node id="zzz" uname="c001n03" type="normal"/>
      </nodes>
      <resources>
        <primitive id="myAddr" class="ocf" provider="heartbeat" type="IPaddr">
          <operations>
           <op id="myAddr-monitor" name="monitor" interval="300s"/>
          </operations>
          <instance_attributes>
             <nvpair name="ip" value="10.0.200.30"/>
          </instance_attributes>
        </primitive>
        <group id="myGroup">
         <primitive id="database" class="lsb" type="oracle">
            <operations>
             <op id="database-monitor" name="monitor" interval="300s"/>
            </operations>
          </primitive>
         <primitive id="webserver" class="lsb" type="apache">
            <operations>
             <op id="webserver-monitor" name="monitor" interval="300s"/>
            </operations>
          </primitive>
        </group>
        <clone id="STONITH">
          <meta_attributes id="stonith-options">
              <nvpair id="stonith-option-1" name="globally-unique" value="false"/>
          </meta_attributes>
          <primitive id="stonithclone" class="stonith" type="external/ssh">
            <operations>
              <op id="stonith-op-mon" name="monitor" interval="5s"/>
            </operations>
            <instance_attributes id="stonith-attrs">
              <nvpair id="stonith-attr-1" name="hostlist" value="c001n01,c001n02"/>
             </instance_attributes>
          </primitive>
        </clone>
     </resources>
      <constraints>
       <rsc_location id="myAddr-prefer" rsc="myAddr" node="c001n01" score="INFINITY"/>
       <rsc_colocation id="group-with-ip" rsc="myGroup" with-rsc="myAddr" score="INFINITY"/>
      </constraints>
    </configuration>
    <status/>
  </cib>


Further Reading

Revision History

Revision History
Revision 119 Oct 2009Andrew Beekhof
Import from Pages.app
Revision 226 Oct 2009Andrew Beekhof
Cleanup and reformatting of docbook xml complete
Revision 3Tue Nov 12 2009Andrew Beekhof
Split book into chapters and pass validation
Re-organize book for use with Publican

Index

F

feedback
contact information for this manual, We Need Feedback!