FLINK-34991 <https://issues.apache.org/jira/browse/FLINK-34991>
When there is a code error that crashes a job manager before the DAG is built, the real error (e.g. ClassNotFoundException) is not propagated correctly. Instead, the CRD shows: "HA metadata not available to restore from last state. It is possible that the job has finished or terminally failed, or the configmaps have been deleted." We initially thought this was a bug that could be fixed in the flink-kubernetes-operator, but it seems there are a couple things in Flink core that prevent the operator from properly exposing the error. Welcome any comments! Thank you, Luca
