This solves #53 . Currently, if a job has finished (either error or not) and at the same time, the shuffle manager leader changes, the new leader notification may cause a fatal error which may lead to JM failover. It mainly influence session cluster when more than one jobs share the same cluster.
Brief change log
Not trigger fatal error if Flink job has finished when notifying new shuffle manager address.
What is the purpose of the change
This solves #53 . Currently, if a job has finished (either error or not) and at the same time, the shuffle manager leader changes, the new leader notification may cause a fatal error which may lead to JM failover. It mainly influence session cluster when more than one jobs share the same cluster.
Brief change log
Verifying this change
This change added tests.