Cheat Sheet

Corosync handles membership and quorum, Pacemaker decides where resources run, and resource agents start, stop and monitor the actual services. pcs configures all of it.

flowchart TB
  pcs[pcs] --> pacemaker[Pacemaker]
  pacemaker -->|start, stop, monitor| agents[Resource agents]
  pacemaker -->|on failure| fence[Fence agent]
  pacemaker <--> corosync[Corosync]
  corosync <-->|membership, quorum| others[Other nodes]
  fence -.->|power off| others

Commands below are for pcs 0.11 (RHEL 9 and later). Older versions use show instead of config and pcs cluster standby instead of pcs node standby.

Status

Command	Shows
pcs status	Nodes, resources, failed actions and daemon status.
pcs status --full	Additionally node attributes and fail counts.
pcs cluster status	Only nodes and the pcsd daemons.
pcs status nodes	Which nodes are online, standby or offline.
crm_mon	Live view that refreshes; crm_mon -1r prints once, including stopped resources.
pcs config	The complete configuration.
pcs cluster cib	The raw CIB (cluster information base) as XML.

Cluster and nodes

Command	Effect
pcs cluster start --all	Starts Corosync and Pacemaker on all nodes (omit --all for the local node).
pcs cluster stop --all	Stops the cluster on all nodes.
pcs cluster enable --all	Starts the cluster at boot.
pcs node standby <node>	Moves all resources off the node, e.g. for maintenance.
pcs node unstandby <node>	Allows resources on the node again.
pcs cluster sync	Copies corosync.conf to all nodes.
pcs cluster corosync	Prints corosync.conf.
pcs cluster destroy	Removes the cluster configuration from the node. Irreversible.

Resources

Command	Effect
pcs resource list	Available resource agents.
pcs resource describe ocf:heartbeat:IPaddr2	Options of an agent.
pcs resource create vip ocf:heartbeat:IPaddr2 ip=10.0.0.10 cidr_netmask=24	Creates a resource.
pcs resource config [<resource>]	Configuration of one or all resources.
pcs resource update <resource> <option>=<value>	Changes options.
pcs resource enable / disable <resource>	Starts or stops a resource, cluster-wide.
pcs resource restart <resource>	Restarts it where it runs.
pcs resource move <resource> <node>	Moves it to another node.
pcs resource cleanup <resource>	Clears failures so Pacemaker tries again.
pcs resource debug-start <resource> --full	Runs the agent outside the cluster and shows its output.
pcs resource group list	Resource groups and their members.
pcs resource delete <resource>	Deletes a resource.
pcs resource defaults update resource-stickiness=100	Makes resources prefer to stay where they are.
pcs constraint config	Location, order and colocation constraints.

Caution

Before pcs 0.11, pcs resource move worked by adding a location constraint that stayed after the move and pinned the resource to that node. Remove it with pcs resource clear <resource>. Newer versions remove it automatically.

Fencing

Command	Effect
pcs stonith list	Available fence agents.
pcs stonith describe <agent>	Options of a fence agent.
pcs stonith create <name> <agent> <options>	Creates a fence device.
pcs stonith config	Configured fence devices.
pcs stonith update <name> <options>	Changes a fence device.
pcs stonith fence <node>	Fences a node, to test the setup.
pcs stonith cleanup	Clears fencing failures.
pcs stonith delete <name>	Deletes a fence device.

Cluster properties

Command	Effect
pcs property config --all	All properties including defaults.
pcs property set stonith-enabled=false	Runs without fencing.
pcs property set no-quorum-policy=ignore	Keeps running resources when quorum is lost.

Danger

Disabling fencing or ignoring quorum is only acceptable in test setups. Without them, both halves of a split cluster can start the same resources and, for example, mount and corrupt the same filesystem. For two-node clusters, Corosync’s two_node: 1 (set automatically by pcs) is the safe way to keep quorum.

Files

Path	Content
/etc/corosync/corosync.conf	Cluster nodes and the communication links.
/var/lib/pacemaker/cib/cib.xml	The cluster configuration. Don’t edit it directly; use pcs.
/etc/sysconfig/pacemaker	Daemon options.
/var/log/pacemaker/pacemaker.log	Pacemaker log.
/var/log/cluster/corosync.log	Corosync log.
/usr/lib/ocf/resource.d/	OCF resource agents, by provider (e.g. heartbeat).