Flink application is running/alive always，but the job is gone

Question

I am using the detach and yarn-cluster mode to run the flink application in job mode as follows:

flink run -d -m yarn-cluster  -yn 10 -ys 1 -yqu QueueA -c com.me.MyFlinkApplicaiton

The application starts up and the job in this application starts to consume message from Kafka successuflly. After running smoothly for serverl hours, the flink yarn application is alive/running, but the job in this application disappears(there is no job/task running any more), all the slots are freed.

My application is a simple read from Kafka source -> sink to mongodb application, and I have try/catch the whole sink function's invoke method, so there will no exception throws in sink function.

I didn't find usefull log to investigate this problem ,so I would ask what may happen that may cause this behavior

Tom Tom · Accepted Answer · 2020-05-11T03:26:22

Ok, looks I have found out the problem, I have specified the restart strategy in the code as

  env.setRestartStrategy(RestartStrategies.noRestart())

When the tm exits and jm is cancelled, flink will not try to restart the jm and tm.

Flink application is running/alive always，but the job is gone

1 Answers