[FLINK-31370] Cancellation of the StreamTask should prevent more timers from being fired

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.18.0
• Component/s: Runtime / Task
• Labels: pull-request-available
• Assignee: David Morávek
• Reporter: David Morávek
• Created: 08/Mar/23 13:54
• Updated: 09/Mar/23 17:36
• Resolved: 09/Mar/23 13:10

Description

If the task is canceled while the watermark progresses, it may be stuck in the Cancelling state for a long time (e.g., when many windows are firing). This is closely related to FLINK-20217, which might bring a more robust solution for checkpoint and cancellation code paths.

As a stopgap solution, we'll introduce a check allowing InternalTimerService to break out of the firing loop if the StreamTask has been marked as canceled.

Issue Links

• causes: FLINK-31387 StreamTaskCancellationTest.testCancelTaskShouldPreventAdditionalProcessingTimeTimersFromBeingFired failed with an assertion
• relates to: FLINK-20217 More fine-grained timer processing
• links to: GitHub Pull Request #22134