Ralph Wadner Blaise
06/17/2024, 7:00 PMjava.util.concurrent.CompletionException: org.apache.flink.runtime.rest.handler.RestHandlerException: Jar file /opt/flink/web-upload/flink-web-upload/xxx.jar does not exist
My Flink Session jobs are getting their jar files from an external Artifactory. I check the network connectivity, and Artifactory and they are both fine. Flink is eventually able to download the jar file. However, when that occurs, it ends up creating duplicate jobs or sometimes more. To be clear, if I deploy 6 jobs, it will create duplicates or more for some because of the jar does not exist error, but all the rest will deploy correctly [single instance]. Is there any way to mitigate this duplicate issue?
P.S The server is running on Kubernetes in HA with JobManager replica of 2 [Standby JobManager]. My suspicion is this could be related to HA and intermittent leader election issues but don't know how to fix it.
Current Tweak High Availability settings:
high-availability.kubernetes.leader-election.lease-duration: 60s
high-availability.kubernetes.leader-election.renew-deadline: 60s
high-availability.kubernetes.leader-election.retry-period: 10s