wombatu-kun commented on code in PR #19514:
URL: https://github.com/apache/hudi/pull/19514#discussion_r3717550973


##########
.github/workflows/bot.yml:
##########
@@ -32,6 +32,13 @@ env:
   FLINK_IT_FILTER2: -Dit.test=!ITTestHoodieDataSource
 
 jobs:
+  # [CI-TRIM] to revisit for CI improvement
+  validate-ci-baseline:

Review Comment:
   Following up on the `validate-ci-baseline` thread with a number, since it 
also changes the rebase ask in #19515.
   
   Because the baseline job sits in `bot.yml`, a pre-trim PR cannot pick up the 
new context without triggering the whole workflow. I measured a recent full run 
at 2,610 runner-minutes (43.5 h); summing the same per-job times over the 17 
surviving matrix entries gives an estimated ~10 h per run. So reviving one 
pre-trim PR costs roughly that, for a required check that is a single `echo`. 
There are 506 open PRs, which puts the full sweep at ~5,000 runner-hours, above 
the 3,600 h five-day limit on its own. Combined with the point above, the job 
currently pays the cost of the test jobs without gating on any of them.
   
   Would it work to move `validate-ci-baseline` into its own small workflow? A 
PR would pick up the context in seconds, and the heavy matrix would run when 
someone actually pushes to it. That would also drop the rebase requirement for 
most PRs: `pull_request` here defaults to `[opened, synchronize, reopened]`, so 
a close/reopen re-reports on the same head SHA, and with `strict: false` and 
GHA building `refs/pull/N/merge` against current master, a rebase adds no 
signal over that except where the branch genuinely conflicts.



-- 
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]

Reply via email to