Hi Yeah good idea to drop the JDK25 then that could allow to run another job instead as those takes the longest. The jenkins CI does a full build and will catch any weird issues that may come on JDK25.
On Thu, Oct 22, 2026 at 9:39 AM Aurelien Pupier via dev < [email protected]> wrote: > Hello, > > I fear that these jobs are not taking that many resources as shown by this > graph: > https://infra-reports.apache.org/#ghactions&project=camel&hours=168&limit=20&group=name > > > The 2 "sides" jobs taking most of the time are PR doc validation (6.81%) > and PR license checks (3.67%). > I think that the PR doc validation is important, when changes are breaking > the website it costs more to detect and rebuild what is needed. For the PR > license check, given that it is checking only the top-level, we could > deactivate for now, or investigate why it is taking so much time just for > that. > > Now that we have a much faster and stable Jenkins builds, I think we could > stop testing with JDK 25 on GitHub Actions always. If we can trigger it > manually (with a comment or a label) for some specific changes this could > be safe enough. The advantage is that it will cut the usage in an important > way, potential reduction of 22.6% just with this change. > > Also note that the whole pool is shared among all Apache projects. Our > usage is important but not that much significant, improvements on our side > alone won't impact much on the queue time. See for instance > https://the-asf.slack.com/archives/CBX4TSBQ8/p1792608273652369?thread_ts=1792520326.623569&cid=CBX4TSBQ8 > > Other options to look at to improve current state: > > - Run PR jobs on user fork by default > - Investigate to fund self-hosted runners (so means doing some try > with different machine size and compare cost/time gain and security > analysis) > - Investigate to use Apache Jenkins nodes which seems slightly less > crowded (but not far from being full from what I observed) > - Investigate using Windows Ci runners which are on another shared > pool which doesn't seem to be at max capacity currently > - Work at Apache org level to monitor and help other high consumers to > reduce their usage > - A bit selfish and temporary workaround: avoid usage of ubuntu-latest > and precise the name to be part of another queue > https://the-asf.slack.com/archives/C0B835NQG6P/p1792614064793079 > > > > regards, > > > ------------------------------ > *From:* Claus Ibsen <[email protected]> > *Sent:* Wednesday, October 21, 2026 8:38 PM > *To:* dev <[email protected]> > *Subject:* [EXTERNAL] [IMPORTANT] - Reduce CI GH actions jobs > > Hi > > The GH actions are dead slow since its overloaded with jobs that stay > queded for many hours or even until next day. It hurt us here at Camel. > > I noticed in GH actions we have some small jobs that does pr-comments > pr-label and what else. > > So in light of this then lets disable as much as possible so we only run GH > jobs that does what we need right now - to build and test the PR and > nothing else. > > Can folks that have knowledge and built those GH actions come forward and > help with this. > > We have them listed (to my understanding in .github) > > https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_apache_camel_tree_main_.github_actions&d=DwIBaQ&c=BSDicqBQBDjDI9RkVyTcHQ&r=0vLfrqGC45Y6-8J0B2MbyQ&m=pQcU-eLBd1D_ihHpe2u2vaeR4ju_HCV_7mOoBEoKJrkaYweCD2lhnl2wlnzyHuxa&s=0xBCldAAAAaUqbDoVLAIpk--Iwf4Vg4PB9PWDn8k4R0&e= > > > > > -- > Claus Ibsen > Unless otherwise stated above: > > Compagnie IBM France > Siège Social : 17, avenue de l'Europe, 92275 Bois-Colombes Cedex > RCS Nanterre 552 118 465 > Forme Sociale : S.A.S. > Capital Social : 664 614 175,50 € > SIRET : 552 118 465 03644 - Code NAF 6203Z > -- Claus Ibsen
