I have an Init Container that does a database version check. If the database version check in the Init Container fails, the whole Deployment should fail immediately. But I still want the other Containers in the Pod to restartPolicy=Always.
However, Init Containers are
retried according to the Pod
restartPolicy
It would be good if I could specify a different restartPolicy for the Init Container than the Deployment but that's not possible. Lifecycle Hooks wouldn't really work either as what you need for this use case is a PreStart hook.
Is there a good way to get Init Containers to fail fast?
Would you like to spin a separate container, which will be checking db version and if succeeds, creates some artifact, like file /tmp/db_version on the host's drive and then you just add livenessProbe/readinessProbe to your main Deployment, which will be checking existence of the /tmp/db_version, e.g.
livenessProbe:
exec:
command:
- ls
- /tmp/db_version
initialDelaySeconds: 5
periodSeconds: 5Thus you'll be able to sideline restartPolicy of the db_version_check container and your main container.