How to set Pacemaker clone resource options

Clone resource options control how many instances of a service can run at once and how those instances are scheduled across the cluster. Correct limits prevent accidental overcommitment while still allowing active-active designs and scale-out workloads to use all available nodes.

In Pacemaker, a clone is a wrapper around an existing resource that tells the scheduler to run multiple instances of that resource. Clone behavior is tuned with meta attributes on the clone definition, including clone-max (cluster-wide instance limit), clone-node-max (per-node instance limit), plus scheduling modifiers such as ordered and interleave that affect start/stop sequencing and dependency handling.

Changing clone meta attributes can trigger rescheduling and restarts as the cluster converges on the new policy. Keep clone-max aligned with the number of eligible nodes intended to run the service, and treat clone-node-max greater than 1 as a deliberate choice that may require per-instance ports, storage isolation, or other safeguards.

1. Show the current clone definition with its meta attributes.

   $ sudo pcs resource config dummy-check-clone
   Clone: dummy-check-clone
     Meta Attributes: dummy-check-clone-meta_attributes
       clone-max=2
     Resource: dummy-check (class=ocf provider=pacemaker type=Dummy)
   ##### snipped #####

   Clone options live on the clone wrapper (dummy-check-clone), not the underlying resource (dummy-check).

   Related: How to show Pacemaker resource configuration

2. Update cluster-wide clone-max plus per-node clone-node-max on the clone resource.

   $ sudo pcs resource update dummy-check-clone meta clone-max=3 clone-node-max=1

   Set clone-max to the maximum number of nodes intended to run the service at the same time.

   Setting clone-max higher than available capacity can start too many instances and overload nodes or exhaust shared dependencies.

3. Set clone scheduling modifiers when ordering or dependency behavior needs adjustment.

   $ sudo pcs resource update dummy-check-clone meta interleave=true ordered=false

   ordered forces sequential start/stop across clone instances, while interleave allows ordered dependent clones to start in parallel per instance.

4. Confirm the clone meta attributes were updated in the cluster configuration.

   $ sudo pcs resource config dummy-check-clone
   Clone: dummy-check-clone
     Meta Attributes: dummy-check-clone-meta_attributes
       clone-max=3
       clone-node-max=1
       interleave=true
       ordered=false
     Resource: dummy-check (class=ocf provider=pacemaker type=Dummy)
   ##### snipped #####

5. Verify the running clone instance count matches the configured limits.

   $ sudo pcs status --full
   ##### snipped #####
   Full List of Resources:
     * Clone Set: dummy-check-clone [dummy-check]:
       * dummy-check (ocf:pacemaker:Dummy): Started node-03
       * dummy-check (ocf:pacemaker:Dummy): Started node-01
       * dummy-check (ocf:pacemaker:Dummy): Started node-02
   ##### snipped #####