SEZ9 commented on PR #11602: URL: https://github.com/apache/seatunnel/pull/11602#issuecomment-5724235381
Thanks @tomatotomata for the thorough baseline work. Reproducing the `CANCELED`-expected/`FAILED`-actual result on unmodified `dev` at `6ee0c3744` as well as on the PR head `a283df0d5`, opening a tracking issue, and narrowing the change to skipping just that one E2E case with the link is exactly the handling I'd hoped for — it keeps the metrics diff untouched and doesn't hide the underlying terminal-state behavior. Noting the baseline result in the PR description is a nice touch too. On the engine-side substance, the departed-worker fix still reads correctly to me: treating a member that has already left the cluster as unavailable and moving on, rather than waiting on it, is the right behavior for the bounded realtime path and doesn't change the strict semantics of the terminal path. I have nothing further open there. On CI: run `35145655123` against `c30188e0fb70` showing both `engine-v2-it` legs passing (`105070776283` and `105070776280`) is the evidence I wanted for this head. Since this PR doesn't touch any connector module, I don't think a closer comparison of the remaining connector jobs against `dev` is needed on your side. One remaining ask before merge: please rebase onto current `dev` so the CI evidence reflects the head we actually merge, and confirm afterwards that the `engine-v2-it` legs are still green with only that single linked skip. Once that's in, I'm happy to move this forward. <!-- streview-comment:1131 --> -- 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]
