CrunchyData
postgres-repo-host pod in CrashLoopBackOff - "environment: /dev/fd/62: Text file busy"
We have a 3 node kubernetes cluster and deployed these examples in a namespace. ( PGO was deployed in namespace and not clusterwide )
This was deployed using steps shared under : https://access.crunchydata.com/documentation/postgres-operator/5.0.0/quickstart/
- We have tried cluster wide operator and namespace based too ( using kustomize ) . In both cases this issue was seen.
Steps : After installation ,restart only "postgres-repo-host-0" pod and its enters CrashLoopBackOff state.
Observed that the postgres-repo-host-0 pod goes into CrashLoopBackOff with container 'pgbackrest-config' shows logs as "environment: /dev/fd/62: Text file busy"
#kubectl logs postgres-repo-host-0 pgbackrest-config -n postgres-operator environment: /dev/fd/62: Text file busy
The pod postgres-repo-host-0 had to be restarted a few times to come back to running state.
This was observed on k8s v1.19 and PGO v 5.0.0, 5.1.0 and also latest ones.
Its seems the monitor function - pgbackrest-config container - is generating this error at step - exec {fd}<> <(:)
Can you please suggest any solution for this issue?
Can anyone please suggest if I can add delay/sleep in or before the monitor function for the pod postgres-repo-host-0 and container pgbackrest-config as the PGO cleans it up after pod restart.
Since you opened up a duplicate issue on the postgres-operator repo (where I think this issue probably belongs), I'm going to close this one and comment over there.
