[ClusterLabs] Systemd resources stopping issue

[Marek Pastierik]

Dear ClusterLabs team,

I would like to ask for clarification regarding Pacemaker behavior with
systemd services.

I am running *Pacemaker 2.1.10-2.el9* on a *3-node cluster* and currently
have *8 cluster resources*, two of which are systemd-based services:
*nginx.service* and *puppetserver.service*.
Both of these services are intended to run on only one node, but I am
testing whether Pacemaker will stop them if I manually start them on a
second node.

I am seeing different behavior for two very similarly defined primitives.

Specifically, *nginx.service* is actively being stopped by Pacemaker on
nodes where it should not be running, while *puppetserver.service* is not,
even though I would expect the same behavior.

For nginx, I can see the following message in pacemaker-controld logs:

pacemaker-controld[520998]: notice: Requesting local execution of stop
operation for nginx_service on X.Y.Z

For puppetserver, I do not see a corresponding stop operation being
requested.

The primitive definitions are:

<primitive id="nginx_service" class="systemd" type="nginx.service">
  <operations>
    <op name="monitor" interval="10s" role="Started"
        id="nginx_service-monitor-interval-10s"/>
    <op name="monitor" interval="11s" role="Stopped"
        id="nginx_service-monitor-interval-11s"/>
    <op name="start" interval="0s"
        id="nginx_service-start-interval-0s"/>
    <op name="stop" interval="0s"
        id="nginx_service-stop-interval-0s"/>
  </operations>
  <meta_attributes id="nginx_service-meta_attributes"/>
</primitive>

<primitive id="puppetserver_service" class="systemd"
type="puppetserver.service">
  <operations>
    <op name="monitor" interval="10s" role="Started"
        id="puppetserver_service-monitor-interval-10s"/>
    <op name="monitor" interval="30s" role="Stopped" timeout="30s"
        id="puppetserver_service-monitor-interval-11s"/>
    <op name="start" interval="0s"
        id="puppetserver_service-start-interval-0s"/>
    <op name="stop" interval="0s"
        id="puppetserver_service-stop-interval-0s"/>
  </operations>
  <meta_attributes id="puppetserver_service-meta_attributes"/>
</primitive>

Could someone please explain why Pacemaker requests a local stop operation
for nginx.service, but not for puppetserver.service?
Is this related to differences in monitor intervals, systemd unit behavior,
or Pacemaker’s internal state handling?

Thank you in advance for any insight.

Best regards,
Marek Pastierik
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/users/attachments/20260121/4fd29415/attachment.htm>