[FLINK-12285] Memory leak in SavepointITCase and SavepointMigrationTestBase

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 1.9.0
• Component/s: Runtime / Coordination, Tests
• Labels: pull-request-available test-stability
• Assignee: Biao Liu
• Reporter: Xiaogang Shi
• Created: 22/Apr/19 06:50
• Updated: 15/May/19 08:33
• Resolved: 15/May/19 08:33

Description

The tests in SavepointITCase and SavepointMigrationTestBase do not cancel running jobs before exit. It will cause exceptions in {{TaskExecutor}}s and unreleased memory segments. Succeeding tests may fail due to insufficient amount of memory.

The problem is caused by cancelling {{TaskExecutor}}s with running tasks. Another issue caused by the reason can be seen in FLINK-11343. Maybe we can find a more dedicated method to cancel those {{TaskExecutor}}s still having running tasks.

Issue Links

• is caused by: FLINK-11630 TaskExecutor does not wait for Task termination when terminating itself
• relates to: FLINK-12332 Cancel running tasks if exist before shutting down TM
• links to: GitHub Pull Request #8269