-
Bug
-
Resolution: Unresolved
-
Normal
-
4.19
-
None
-
Quality / Stability / Reliability
-
False
-
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
-
None
Description of problem:
See this line: https://github.com/openshift/cluster-etcd-operator/blob/main/pkg/tnf/pkg/pcs/etcd.go#L25
For some reason, possibly a bug in pacemaker, initializing the podman-etcd resource agent in with `meta notify=true` doesn't set the OCF_RESKEY_CRM_meta_notify flag like it's supposed to. In our testing, it has been possible to override this this by omitting "meta" in the resource creation step.
How reproducible:
Run the graceful recovery test on a TNF cluster. OR Run `crm_mon --as-xml` on one of the CP nodes
Steps to Reproduce:
1. Deploy a TNF cluster 2. Try to recover from graceful shutdown 3. Cluster doesn't recover, so observe the logs in the CP node
Actual results:
Cluster doesn't recover
Expected results:
Cluster recovers
Additional info:
None
- blocks
-
OCPBUGS-55691 TNF: initializating podman-etcd with meta causes OCF_RESKEY_CRM_meta_notify to be set to false
-
- Closed
-
- is cloned by
-
OCPBUGS-55691 TNF: initializating podman-etcd with meta causes OCF_RESKEY_CRM_meta_notify to be set to false
-
- Closed
-
- links to