In the current implementation of the subreconcilers a Pod that is in a Pending state or is unreachable can delay the selection of new coordinators. The reason is that UpdatePodConfig will trigger a requeue if not all Pods are reachable and have to correct ConfigMap. Depending how long the Pod is in that state this can be critical for the cluster since a coordinator that's lost during that time will not be replaced by a new one which could lead to losing the quorum.
In the current implementation of the subreconcilers a Pod that is in a Pending state or is unreachable can delay the selection of new coordinators. The reason is that
UpdatePodConfig
will trigger a requeue if not all Pods are reachable and have to correct ConfigMap. Depending how long the Pod is in that state this can be critical for the cluster since a coordinator that's lost during that time will not be replaced by a new one which could lead to losing the quorum.We have a related issue with a different focus: https://github.com/FoundationDB/fdb-kubernetes-operator/issues/732