[KAFKA-12738] Improved error handling for better at-least-once semantics and faster EOS

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: streams
• Labels: new-streams-runtime-should-fix
• Assignee: Unassigned
• Reporter: A. Sophie Blee-Goldman
• Created: 01/May/21 01:19
• Updated: 19/Apr/22 09:41

Description

Umbrella ticket for various ideas I had to improve the error handling behavior of Streams

Issue Links

• links to: GitHub Pull Request #11787
• links to: GitHub Pull Request #11801
• links to: GitHub Pull Request #11812
• links to: GitHub Pull Request #12044

Sub-Tasks

• 1. Commit all healthy tasks after a task-specific error for better task isolation: Open
• 2. Commit any cleanly-processed records within a corrupted task: Open
• 3. Resume processing from last-cleared processor after soft crash: Open
• 4. Handle thread-wide errors: Open
• 5. Checkpoint all uncorrupted state stores within the subtopology: Open
• Allow StreamThreads to recover from some exceptions without killing/replacing the thread: Open