DanielLeens commented on issue #10995: URL: https://github.com/apache/seatunnel/issues/10995#issuecomment-4655019822
This is worth treating as a small CI policy / process issue, not just a one-off workflow cleanup. I agree with the general tiered direction already discussed above, and I would keep the first deliverable intentionally small: 1. document the split thresholds clearly 2. document the expected edit points (`backend.yml`, `update_modules_check.py`, and any checklist entry) 3. define one ownership rule: contributors can propose the split, but a CI-aware maintainer should review it 4. add one machine-checkable guard where possible so we do not silently drop or double-count modules For the thresholds themselves, the current direction looks reasonable: - standalone job when one module is consistently > 60 minutes - rebalance grouped jobs when a group drifts past roughly 90 minutes - reconsider the parent layout when the aggregated path drifts past roughly 2.5 hours On the UT / IT fragility point: I would be careful not to overload the first version of the rule with every CI pain at once. Duration and queue pressure are easy to observe and standardize. General flakiness is much harder to score consistently, and mixing both goals immediately may slow down the part we already know how to improve. So my suggestion would be: - phase 1: standardize duration-based splitting and the review checklist - phase 2: if needed, add a separate rule or signal for flaky / environment-sensitive modules If someone wants to turn this into a concrete follow-up PR, I think the best first PR is docs + checklist + one guard improvement, rather than trying to redesign all CI grouping rules in a single shot. -- 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]
