zhangshenghang commented on PR #11500: URL: https://github.com/apache/seatunnel/pull/11500#issuecomment-6340208767
Issue 1 is addressed in 0961519e9 (en+zh incompatible-changes): the migration sizing no longer presents `N = 2 + Σ job parallelism` as the universal formula. It now counts coordinator and physical task groups per concurrent job — roughly `Σ (2 + P_j)` per job in the simple one-enumerator/one-committer case (two independent P=1 jobs need about six slots, not four) — and explicitly notes that sinks without an aggregate committer and multi-source/multi-pipeline plans need more, and that an aggregate count alone is not a placement guarantee. Issue 2 needs a decision we should not make unilaterally: the factory-inventory vs global-inventory fallback in `ContainerTestingExtension` is shared test infrastructure, and your review offers two legitimate paths (fix the fallback to use the suite factory's own pre/post-selection inventory here, or move the shared-infrastructure correction to its owning work #12895). Since @SEZ9 accepted the narrowed condition while you showed a source-derived counterexample, we'd rather not silently broaden skip policy inside a default-slot PR. Please advise which path you prefer — if it is 'fix here', we will implement the factory-scoped fallback plus injected-discovery unit coverage; if it is 'move to #12895', we will revert the shared-extension part of the diff from this PR and keep the default-flip scope. -- 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]
