[FLINK-23201] The check on alignmentDurationNanos seems to be too strict

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.12.2, 1.14.0, 1.13.1
• Fix Version/s: 1.14.0, 1.12.5, 1.13.2
• Component/s: Runtime / Metrics
• Labels: pull-request-available
• Assignee: Anton Kalashnikov
• Reporter: Jun Qin
• Created: 01/Jul/21 07:33
• Updated: 23/Sep/21 18:04
• Resolved: 15/Jul/21 17:53

Description

The check on alignmentDurationNanos seems to be too strict at the line:
https://github.com/apache/flink/blob/master/flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/CheckpointMetrics.java#L74

This caused a job to fail when doing stop-with-savepoint. But doing savepoint only without stop does not seem to be impacted by this.

Issue Links

• causes: BEAM-10955 testSavepointRestoreLegacy flake: Could not find Flink job (FlinkJobNotFoundException)
• is duplicated by: FLINK-23302 Precondition failed building CheckpointMetrics.
• links to: GitHub Pull Request #16361