[FLINK-20413] Sources should add splits back in "resetSubtask()", rather than in "subtaskFailed()".

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.11.3, 1.12.0
• Component/s: Runtime / Coordination
• Assignee: Stephan Ewen
• Reporter: Stephan Ewen
• Created: 29/Nov/20 21:53
• Updated: 12/Dec/20 11:58
• Resolved: 30/Nov/20 17:48

Description

Because "subtaskFailed()" has no strong order guarantees with checkpoint completion, we need to return failed splits in "resetSubtask()" instead.

See FLINK-20396 for a detailed explanation of the race condition.

Issue Links

• is blocked by: FLINK-20396 Add "OperatorCoordinator.resetSubtask()" to fix order problems of "subtaskFailed()"