[FLINK-26592] [Changelog] Deadlock in FsStateChangelogWriter

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.15.0
• Fix Version/s: 1.15.0, 1.16.0
• Component/s: Runtime / State Backends
• Labels: pull-request-available
• Assignee: Roman Khachatryan
• Reporter: Roman Khachatryan
• Created: 10/Mar/22 22:15
• Updated: 21/Mar/22 11:42
• Resolved: 21/Mar/22 11:41

Description

The issue occurs when sizes of buffers are set to minimum (e.g. 1 byte).
Task thread tries to update state -> schedules to upload changes -> waits for capacity.
Upload threads do release capacity on upload completion; however, they are unable to send back the results because the Writer lock is taken; therefore, they're unable to proceed with the next uploads.

Issue Links

• Discovered while testing: FLINK-25144 Test Changelog State Backend manually
• links to: GitHub Pull Request #19050
• links to: GitHub Pull Request #19062