Pacemaker architecture - the resource manager

Pacemaker is the resource manager for a Corosync cluster. It decides which node runs which resource, monitors the resources, and triggers failover when needed.

What Pacemaker does

Pacemaker sits on top of Corosync:

• Decides which node runs which resource.
• Monitors resource health.
• Fails over when a resource fails.
• Enforces constraints (e.g. “DB must be on a different node than the application”).

Pacemaker does not know how to manage a specific service (e.g. PostgreSQL). It uses resource agents (RAs) for that.

Resource agents

An RA is a script (or compiled binary) that knows how to manage a service:

• start: start the service.
• stop: stop the service.
• monitor: check if the service is running.
• promote / demote: for master/slave resources.
• migrate: for live migration.

Standard RAs are shipped with the resource-agents package and cover most services (PostgreSQL, MySQL, nginx, etc.). For custom services, write a custom RA.

Configure a resource

# Apache
pcs resource create web apache \
    configfile=/etc/apache2/apache2.conf \
    statusurl=http://localhost/server-status

# PostgreSQL
pcs resource create db pgsql \
    pgctl=/usr/bin/pg_ctl \
    psql=/usr/bin/psql \
    pgdata=/var/lib/pgsql/data \
    op monitor interval=30s

# IP address
pcs resource create vip ocf:heartbeat:IPaddr2 \
    ip=10.0.0.100 \
    op monitor interval=30s

Pacemaker creates the resource, monitors it, and starts it on a node.

Constraints

Constraints tell Pacemaker where resources can run:

# Location: web must run on node1 or node2
pcs constraint location web prefers node1
pcs constraint location web prefers node2

# Colocation: db must run on the same node as web
pcs constraint colocation add web with db

# Order: db starts before web
pcs constraint order db then web

Three types:

• Location: where a resource can run.
• Colocation: which resources run together.
• Order: which resources start/stop first.

Cluster status

pcs status

Output:

Cluster name: mycluster
Stack: corosync
Current DC: node2 (version 2.x) - partition with quorum
Last updated: ...
3 nodes configured
3 resources configured

Online: [ node1 node2 node3 ]

Full list of resources:

  * web    (ocf::heartbeat:apache): Started node2
  * db     (ocf::heartbeat:pgsql):  Master node2
  * vip    (ocf::heartbeat:IPaddr2):  Started node2

The output shows:

• Cluster name and stack.
• Current DC (designated coordinator).
• Quorum status.
• Nodes online.
• Resources and their state.

Failover

When a resource fails:

1. Pacemaker detects via the monitor operation.
2. Pacemaker applies constraints and decides the new node.
3. Pacemaker stops the resource on the failed node.
4. Pacemaker starts the resource on the new node.

If the new node is unreachable, fencing kicks in.

Knowledge check