Pacemaker Role for Ansible

This role sets up a Pacemaker cluster by saving the configuration (CIB), modifying the XML, and reloading it. It is designed to be idempotent, meaning running it multiple times won’t change the result after the first application. It also supports check mode to preview changes.

The role has been updated to manage individual components (like cluster defaults, resources, groups, constraints, etc.) instead of configuring the entire cluster and all services at once. This lets you concentrate on specific resources without affecting everything else.

Requirements

This role is intended for Scientific Linux 7, but it might work on other Linux distributions too; please share any experiences you have.

Tasks

Use the tasks_from Ansible directive to indicate what you want to set up.

In properties, Boolean values interpreted by Pacemaker don’t need quotes. However, resource agents may require these Boolean-like values in formats like integers or strings, and they must be quoted.

tasks_from: main

Set up nodes, configure cluster properties, and resource defaults.

pcmk_cluster_name

Name of the cluster (optional).

Default: hacluster.

pcmk_password

Plaintext password for the cluster user (optional). If not provided, it will be generated from ansible_machine_id of the first host in the play batch. This password is used only for the initial node authentication.

Default: ansible_machine_id | to_uuid

pcmk_user

System user to authenticate PCS nodes (optional). PCS will handle authentication among all nodes.

Default: hacluster

pcmk_cluster_options

A dictionary of cluster-wide options (optional).

pcmk_votequorum

A dictionary of votequorum options (optional). See votequorum(5). Boolean values are accepted.

pcmk_resource_defaults

A dictionary specifying resource defaults (optional).

tasks_from: resource

Set up a simple resource.

pcmk_resource

A dictionary describing a simple (primitive) resource, containing:

• id: mandatory resource identifier for simple resources;
• class, provider, and type: resource agent details; provider can be omitted (e.g., when type is service);
• options: optional dictionary for resource-specific attributes (e.g., address for IPaddr2);
• op: optional list of operations; each operation is a dictionary with required name and interval, plus other optional fields;
• meta: optional dictionary for meta-attributes.

tasks_from: group

Set up a resource group.

pcmk_group

Dictionary with two members:

• id: group identifier;
• resources: a dictionary where keys are resource IDs, and values have formats like pcmk_resource (except the resource id is optional).

tasks_from: constraint

Set up a constraint.

pcmk_constraint

A dictionary defining a single constraint, requiring:

• type: one of: location, colocation, or order;
• score: constraint score (a signed integer, INFINITY, or -INFINITY).

Depending on type, additional required fields are:

• location needs rsc and node;
• colocation needs rsc and with-rsc;
• order needs first and then.

The dictionary may include other fields, like symmetrical.

Example Playbooks

Active-active chrooted BIND DNS Server

---
- name: Configure DNS cluster
  hosts: dns-servers
  tasks:

    - name: Set up cluster
      include_role:
        name: devgateway.pacemaker
      vars:
        pcmk_password: hunter2
        pcmk_cluster_name: named
        pcmk_cluster_options:
          stonith-enabled: false

    - name: Configure IP address resource
      include_role:
        name: devgateway.pacemaker
        tasks_from: resource
      vars:
        pcmk_resource:
          id: dns-ip
          class: ocf
          provider: heartbeat
          type: IPaddr2
          options:
            ip: 10.0.0.1
            cidr_netmask: 8
          op:
            - name: monitor
              interval: 5s

    - name: Configure cloned BIND resource
      include_role:
        name: devgateway.pacemaker
        tasks_from: advanced-resource
      vars:
        pcmk_resource:
          type: clone
          id: dns-clone
          resources:
            named:
              class: service
              type: named-chroot
              op:
                - name: monitor
                  interval: 5s

    - name: Set up constraints
      include_role:
        name: devgateway.pacemaker
        tasks_from: constraint
      vars:
        pcmk_constraint:
          type: order
          first: dns-ip
          then: dns-clone

Active-active Squid Proxy

---
- name: Configure Squid cluster
  hosts: proxy-servers
  tasks:

    - name: Set up cluster
      include_role:
        name: devgateway.pacemaker
      vars:
        pcmk_password: hunter2
        pcmk_cluster_name: squid
        pcmk_cluster_options:
          stonith-enabled: false

    - name: Configure IP address resource
      include_role:
        name: devgateway.pacemaker
        tasks_from: resource
      vars:
        pcmk_resource:
          id: squid-ip
          class: ocf
          provider: heartbeat
          type: IPaddr2
          options:
            ip: 192.168.0.200
            cidr_netmask: 24
          op:
            - name: monitor
              interval: 5s

    - name: Configure cloned Squid resource
      include_role:
        name: devgateway.pacemaker
        tasks_from: advanced-resource
      vars:
        pcmk_resource:
          id: squid
          type: clone
          resources:
            squid-service:
              class: service
              type: squid
              op:
                - name: monitor
                  interval: 5s

    - name: Set up constraints
      include_role:
        name: devgateway.pacemaker
        tasks_from: constraint
      vars:
        pcmk_constraint:
          type: order
          first: squid-ip
          then: squid

Nginx, Web Application, and Master-Slave Postgres

This cluster runs two Postgres nodes with synchronous replication. A virtual IP address follows the master node, and Nginx with a web application run on the same node, conserving resources. This is based on a Clusterlabs wiki example.

---
- hosts:
    - alpha
    - bravo
  tasks:

    - name: Set up Pacemaker with Postgres master/slave
      include_role:
        name: devgateway.pacemaker
      vars:
        pcmk_pretty_xml: true
        pcmk_cluster_name: example
        pcmk_password: hunter2
        pcmk_cluster_options:
          no-quorum-policy: ignore
          stonith-enabled: false
        pcmk_resource_defaults:
          resource-stickiness: INFINITY
          migration-threshold: 1

    - name: Configure simple resources
      include_role:
        name: devgateway.pacemaker
        tasks_from: resource
      loop_control:
        loop_var: pcmk_resource
      loop:
        - id: coolapp
          class: service
          type: coolapp
        - id: nginx
          class: service
          type: nginx
        - id: virtual-ip
          class: ocf
          provider: heartbeat
          type: IPaddr2
          options:
            ip: 10.0.0.23
          meta:
            migration-threshold: 0
          op:
            - name: start
              timeout: 60s
              interval: 0s
              on-fail: restart
            - name: monitor
              timeout: 60s
              interval: 10s
              on-fail: restart
            - name: stop
              timeout: 60s
              interval: 0s
              on-fail: restart

    - name: Configure master-slave Postgres
      include_role:
        name: devgateway.pacemaker
        tasks_from: advanced-resource
      vars:
        pcmk_resource:
          id: postgres
          type: master
          meta:
            master-max: 1
            master-node-max: 1
            clone-max: 2
            clone-node-max: 1
            notify: true
          resources:
            postgres-replica-set:
              class: ocf
              provider: heartbeat
              type: pgsql
              options:
                pgctl: /usr/pgsql-9.4/bin/pg_ctl
                psql: /usr/pgsql-9.4/bin/psql
                pgdata: /var/lib/pgsql/9.4/data
                rep_mode: sync
                node_list: "{{ ansible_play_batch | join(' ') }}"
                restore_command: cp /var/lib/pgsql/9.4/archive/%f %p
                master_ip: 10.0.0.23
                restart_on_promote: "true"
                repuser: replication
              op:
                - name: start
                  timeout: 60s
                  interval: 0s
                  on-fail: restart
                - name: monitor
                  timeout: 60s
                  interval: 4s
                  on-fail: restart
                - name: promote
                  timeout: 60s
                  interval: 0s
                  on-fail: restart

    - name: Set up constraints
      include_role:
        name: devgateway.pacemaker
        tasks_from: constraint
      loop_control:
        loop_var: pcmk_constraint
      loop:
        - type: colocation
          rsc: virtual-ip
          with-rsc: postgres
          with-rsc-role: Master
          score: INFINITY
        - type: colocation
          rsc: nginx
          with-rsc: virtual-ip
          score: INFINITY
        - type: colocation
          rsc: coolapp
          with-rsc: virtual-ip
          score: INFINITY
        - type: order
          first: postgres
          first-action: promote
          then: virtual-ip
          then-action: start
          symmetrical: false
          score: INFINITY
        - type: order
          first: postgres
          first-action: demote
          then: virtual-ip
          then-action: stop
          symmetrical: false
          score: 0

See Also

• The official Pacemaker documentation

Copyright

Copyright 2015-2019, Development Gateway. Licensed under GPL v3+.