Hi Team, We have a data pipeline which is built using Apache Beam SDK <https://beam.apache.org/> and we use Apache Flink Runner <https://beam.apache.org/documentation/runners/flink/> to execute Beam pipelines. We use the local embedded execution mode of Flink for running the pipelines.
Currently, we are running into an issue where in the batch pipeline can be triggered multiple times and each time the pipeline is triggered the Flink creates a miniCluster in local execution mode, runs the job and destroys the miniCluster once the job is completed. When the batch job is triggered multiple times we observe the application process Resident Set Size(RSS) memory keeps increasing (approximately increases by a value set for parameter 'taskmanager.memory.network.max') with each run and is not released, eventually leading to crash of the docker container (container memory is limited) in which this is deployed. However, when we checked the JVM memory using tools like JConsole, jcmd etc does not show any increase and is always with the Xmx value set. We analysed the heap dump of the application but it didn't show any memory leaks. Has anyone faced this issue? Any pointers are appreciated. Thanks, Chandra
