By design, kubernetes environment collect logs from pods and store them temporarily as files on worker nodes. The contents of those files is what is being logged by commands: oc logs <pod> or oc log --previous <pod>

The logs on disk are only kept for the last two interation of a pod. As a corollary, once a pod has restarted three times, old logs are pruned by k8s and the service logs are lost.

This is an issue for investigation purpose, as it is not uncommon that the galera pods are restarted, and we do not enforce any logging for the pods from the OpenStack control plane.