GitHub user mengw15 added a comment to the discussion: Proposal: Automatically terminate idle Kubernetes Computing Units
Yes, the result retrieval path is currently tied to the original live CU (results are fetched through the workflow websocket handled inside the CU), so once the CU is gone the UI can't reach them even if the data is still in storage. I documented this in #5363, we haven't planned to work on it yet. I agree with @kunwp1 that a separate CU-independent retrieval path would be a fairly large change. @kunwp1 On the retention idea — I remember we already have two related mechanisms: (1) a result-retention parameter result-cleanup.ttl-in-seconds (env RESULT_CLEANUP_TTL_IN_SECONDS, default 24h) with a periodic cleanup in ComputingUnitMaster; (2) a ~30s cleanup (web-server.workflow-state-cleanup-in-seconds) that clears the latest execution's results once the last user disconnects and the workflow is no longer running. Since you have more context on this part, does that match your understanding? GitHub link: https://github.com/apache/texera/discussions/6264#discussioncomment-17644424 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected]
