PerconaDB Pods are Restarting Frequently due to OOMKilled Errors

Problem

Frequent PerconaDB pod restarts are observed due to OOMKilled errors. The output of the **percona-db-pxc-db-pxc** pod describe:

Containers:
...      
   pxc:
    Image:         percona_percona-xtradb-cluster:8.0.39-30.1
    Ports:         3306_TCP, 4444_TCP, 4567_TCP, 4568_TCP, 33062_TCP, 33060_TCP
    Host Ports:    0_TCP, 0_TCP, 0_TCP, 0_TCP, 0_TCP, 0_TCP
    Command:
      _var_lib_mysql_pxc-entrypoint.sh
    Args:
      mysqld
    State:          Running
      Started:      Tue, 03 Jun 2025 10:58:46 +0000
    Last State:     Terminated
      Reason:       OOMKilled                            
      Exit Code:    137                              
      Started:      Tue, 03 Jun 2025 06:54:25 +0000
      Finished:     Tue, 03 Jun 2025 10:58:43 +0000
    Ready:          True
    Restart Count:  73                                   
    Limits:
      cpu:     2
      memory:  6Gi                                     
    Requests:
      cpu:      400m
      memory:   3Gi

Environment

• Self-Hosted Private Cloud Director Virtualization - v2025.4 and Higher
• Self-Hosted Private Cloud Director Kubernetes - v2025.4 and Higher

Cause

The current limit for the **percona-db-pxc-db-pxc-<0-2>** pod memory limit is insufficient for its operations.

Diagnostics

Commands used to identify the issue:

$ kubectl describe pod -n <REGION_NAMESPACE> percona-db-pxc-db-pxc-<0-2>
$ kubectl events -n <REGION_NAMESPACE>

Resolution

Increase the memory limit of the pods from the statefulset **percona-db-pxc-db-pxc**

$ kubectl edit statefulset percona-db-pxc-db-pxc-<0-2> -n <REGION_NAMESPACE> 
...
      resources:
        limits:
          cpu: "2"
          memory: 8Gi    # Default value 6Gi
...

Validation

Make sure the pods are running as expected and the new memory limits are updated in the pods

$ kubectl describe pod -n <REGION_NAMESPACE> percona-db-pxc-db-pxc-<0-2>
$ kubectl get pods -n <REGION_NAMESPACE>

Additional Information

If the PerconaDB pods are down, it might impact other pods like the pf9-nginx keystone etc.