How to set Pacemaker quorum policy

Quorum policy determines how a Pacemaker cluster reacts when a partition loses quorum, such as during node failures or network splits. Choosing the right behavior balances availability against the risk of running services without a clear majority.

In most Linux clusters, Corosync tracks membership and quorum while Pacemaker schedules resource actions based on that quorum state. The cluster-wide no-quorum-policy property defines the action to take when a partition is not quorate, and it is stored in the cluster configuration (CIB) managed through pcs.

Relaxed policies like ignore can keep services online but increase split-brain risk when fencing (STONITH) is missing, broken, or slow. Conservative policies like stop or freeze reduce that risk by halting resources or preventing recovery in non-quorate partitions, at the cost of downtime after failures. Two-node clusters are a special case where quorum design (quorum device or policy selection) must be deliberate to avoid both nodes trying to serve the same workload.

1. Display cluster properties to confirm the current no-quorum-policy value.

   $ sudo pcs property config
   Cluster Properties: cib-bootstrap-options
     cluster-infrastructure=corosync
     cluster-name=clustername
     dc-version=2.1.6-6fdc9deea29
     have-watchdog=false
     stonith-enabled=false

   Some pcs versions accept pcs property show no-quorum-policy as a shorter query.

2. Select the target no-quorum-policy value based on the cluster design.

   Common no-quorum-policy values:

   • stop - Stop all resources in a partition without quorum.
   • freeze - Keep current resource state, but avoid recovering resources from nodes outside the partition.
   • ignore - Continue resource management without quorum.

   Some Pacemaker versions also provide values such as demote, fence, or suicide for promotable resources and fencing behavior.

3. Set the no-quorum-policy cluster property to the chosen value.

   $ sudo pcs property set no-quorum-policy=freeze

   Setting no-quorum-policy=ignore can allow split-brain and data corruption when fencing fails or is disabled.

   Restore the default value by clearing the property with sudo pcs property set no-quorum-policy=.

4. Re-check cluster properties for the updated no-quorum-policy value.

   $ sudo pcs property config
   Cluster Properties: cib-bootstrap-options
     cluster-infrastructure=corosync
     cluster-name=clustername
     dc-version=2.1.6-6fdc9deea29
     have-watchdog=false
     no-quorum-policy=freeze
     stonith-enabled=false

5. Check quorum runtime status after the change.

   $ sudo pcs quorum status
   Quorum information
   ------------------
   Date:             Wed Dec 31 12:07:51 2025
   Quorum provider:  corosync_votequorum
   Nodes:            3
   Node ID:          1
   Ring ID:          1.3d
   Quorate:          Yes
   
   Votequorum information
   ----------------------
   Expected votes:   3
   Highest expected: 3
   Total votes:      3
   Quorum:           2
   Flags:            Quorate
   
   Membership information
   ----------------------
       Nodeid      Votes    Qdevice Name
            1          1         NR node-01 (local)
            2          1         NR node-02
            3          1         NR node-03