When performing a kubectl rolling-update
of a replication controller in Kubernetes (Google Container Engine), the Google (Stackdriver) Logging agent doesn't pick up the newly deployed pod. The Log is stuck at the last message produced from the old pod.
Consequently, the logs for the replication controller are out-of-date until we do a manual restart (i.e. kubectl scale
and kubectl delete
) of the pod and the logs are updated again.
Can anybody else confirm that behaviour? Is there a workaround?