[FLINK-17308] ExecutionGraphCache cachedExecutionGraphs not cleanup cause OOM Bug

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.9.0, 1.9.1, 1.9.2, 1.10.0
• Fix Version/s: 1.9.4, 1.10.1, 1.11.0
• Component/s: Runtime / REST
• Labels: pull-request-available
• Assignee: Till Rohrmann
• Reporter: yujunyong
• Created: 21/Apr/20 14:40
• Updated: 30/Apr/20 05:03
• Resolved: 24/Apr/20 13:21

Description

class org.apache.flink.runtime.rest.handler.legacy.ExecutionGraphCache will cache job execution graph in field 

"cachedExecutionGraphs" when call method 

"getExecutionGraph", but never call it's 

cleanup method in flink. it's cause JobManager Out of Memory, When submit a lot of batch job and fetch these job's info. becasue these operation cache all these job execution graph and "cleanup" method never called

Issue Links

• is duplicated by: FLINK-16399 The ExecutionGraphCache class does not periodically clear stale data, resulting in JM memory leaks
• links to: GitHub Pull Request #11879