Dirk Marwinski

> @marwinski Is that a possible issue that may lead to the observed, but not necessarily perceived API server outages? They are phantom outages? Yes, these are the outages shown...

We will enable etcd secret encryption (https://kubernetes.io/docs/tasks/administer-cluster/encrypt-data/) so I believe that having an encrypted etcd EBS volume is not necessary. We need to ensure however that the snapshots and full...

Sorry for the delay contributing here. I cannot really comment on the best way of doing the encryption but need to mention one important aspect: please do not copy the...

We have confirmed that issue 2 above (missing flow control) is indeed an issue although the operating system buffers appear to be slightly bigger than anticipated. We have worked on...

Hi @ydp, this issue could well cause frequent timeouts from kubectl logs as it goes via the tunnel. #167 looks like the error handling issue described in (3) above. You...

Here is some initial investigation: - to the best of my knowledge we only see this issue with Kubernetes 1.21.x which is still in preview. I have checked a 1.19...

No a solution but possibly a viable workaround: Restarting docker removes the stale container, so instead of just restarting the kubelet, docker should be restarted beforehand. Some further analysis. On...

PS: This is the docker issue that describes the problem: https://github.com/docker/for-linux/issues/1148 Unfortunately there is no solution but it mentions the same workaround as I did above.

@MrBatschner: The OS extension is probably the only component that restarts the kubelet!? It appears that docker can be restarted without stopping all the containers (see https://docs.docker.com/config/containers/live-restore/#enable-live-restore). It also appears...

This should run with the integration tests. /cc @MrBatschner