SEZ9 commented on issue #12457: URL: https://github.com/apache/seatunnel/issues/12457#issuecomment-5825994374
Thanks for the detailed STIP. I agree with the direction already given in this thread: keep v1 YARN-first, freeze the YARN lifecycle contract first, and land the implementation in reviewable slices, with Kubernetes following under its own contract. A few things that would help make the STIP reviewable on that basis: 1. Please update the STIP to describe the migration from the existing YarnResourceManager / KubernetesResourceManager placeholders to the proposed resource manager driver, so there is a single owner for external worker allocation, release, and terminal cleanup. 2. On checkpoint recovery: the proposal says a new application restores from the historical job ID and fails if no eligible checkpoint exists. Could you spell out how the restore job ID is resolved against the configured checkpoint storage (path/namespace lookup) and what makes a checkpoint "eligible"? Confirming that this reuses the existing native checkpoint storage without a new mechanism would help. 3. Please state the YARN-first split and the deferral of Kubernetes explicitly in the STIP and in #12460, and link the follow-up work so each slice can be reviewed on its own. Happy to help review once the STIP reflects this. <!-- streview-comment:1307 --> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
