Yes, that’s why we are still pushing the optimization for our CI system even though we are kind of under limit now. We want the bandwidth for more coverage. It should be a temp solution to disable post merge CI on branch-4.x. We should eventually at least run some tests for each commit.
Cheng Pan <[email protected]>于2026年7月28日 周二23:53写道： > +1 to extend the merge script to add a label in the commit message. > > BTW, I feel the current CI skipping strategy might be overaggressive, for > example, branch-4.x doesn’t run CI, and it has been broken for a few days. > If we adopt this proposal, do we have some CI bandwidth to enable CI for > branch-4.x? > > Thanks, > Cheng Pan > > > > On Jul 29, 2026, at 14:33, Wenchen Fan <[email protected]> wrote: > > It's hard to ask committers to manually edit the git message. Shall we > introduce a PR label, and the merge script can adjust the commit message if > it sees the PR label? > > On Wed, Jul 29, 2026 at 9:30 AM Tian Gao via dev <[email protected]> > wrote: > >> https://github.com/apache/spark/actions/metrics/usage >> >> I believe this is accessible to everyone. The ASF github action worker >> policy is at https://infra.apache.org/github-actions-policy.html >> >> On Tue, Jul 28, 2026 at 5:37 PM Nicholas Chammas < >> [email protected]> wrote: >> >>> >>> > On Jul 28, 2026, at 3:26 PM, Tian Gao via dev <[email protected]> >>> wrote: >>> > >>> > We've been struggling with our CI usage. With significant recent >>> effort from the community, we have managed to keep usage barely under the >>> limit (though sometimes slightly over). >>> >>> Is there some place we can see a report on our usage vs. the ASF limit? >>> Do we have a basic analysis of where the CI time is being spent, broken >>> down by job and step? >>> >>> I don’t have a opinion on the git trailer (mainly because it’s a >>> decision for committers only), but as a contributor who has recently spent >>> a good chunk of time studying our CI infrastructure, I have wondered about >>> where our usage is all going. >>> >>> Nick >>> >>> >
