On Wed, 2022-04-20 at 14:24 +0200, Gavin McDonald wrote: > Hi Robert > > On Wed, Apr 20, 2022 at 2:10 PM Robert Munteanu <romb...@apache.org> > wrote: > > > On Wed, 2022-04-20 at 09:37 +0200, Gavin McDonald wrote: > > > On Wed, Apr 20, 2022 at 9:32 AM Robert Munteanu > > > <romb...@apache.org> > > > wrote: > > > > > > > Hi, > > > > > > > > On Wed, 2022-04-20 at 09:27 +0200, Gavin McDonald wrote: > > > > > 64G Sling > > > > > > > > We have large a number of jobs at > > > > > > > > https://ci-builds.apache.org/job/Sling/job/modules/ > > > > > > > > Would it be possible to get a listing of the top jobs in terms > > > > of > > > > disk > > > > usage? > > > > > > > > > > Sure listing those above 100mb > > > > > > 12G sling-org-apache-sling-app-cms > > > > I'm looking at this one since it's the largest and I am not sure > > where > > the large disk usage is coming from. What exactly is retained by > > Jenkins? > > > > We do have a retention policy of 10 builds set globally for Sling > > builds, it can be seen for instance at [1], but maybe that is too > > much > > if we have too many active branches. > > > > There are multiple things we can tweak, but I'd like to see the > > data > > > > /home/jenkins/jenkins-home/jobs/Sling/jobs/modules/jobs/sling-org- > apache-sling-app-cms/branches > # du -sh * | sort -rh > 6.2G master - of which there are 11 jobs stored at 635M each > > 635M PR-14 > 635M PR-13 > 635M dependabot.o0l8t0g19.imist-1-2-6 > 635M dependabot.d9qujnh19.imist-1-2-6 > 626M sitemap-support > 598M PR-12 > 596M dklco-preview-img.ln1s45 > 397M forms-enhancements > 169M SLING-8917-ldap > 163M SLING-8947-grid-missing-fields > 162M SLING-8930-i18n-not-reloading > 162M SLING-8919-fix-error-dialog > 3.9M SLING-11104 > 2.2M SLING-8913-m.h8mc0v.stance-types > 1.7M PR-11 > 1.7M graphql-support > 1.2M dklco-form-framework.7q4mpe > 1012K commons-thumbnails > 860K dklco-conf-cleanup.seumnn > 692K provisioning-support
That is helpful, thank you. I have adjusted the global 'Orphaned Items Strategy' for the Sling modules to only retain old jobs for 3 days. I think that the old limit of 15 jobs was causing PR builds to stick around too long and contributed to the increase in disk usage. This has triggered a rescan of the Sling modules on GitHub [1] which has started cleaning up modules. However, it's going slowly due to GitHub API rate limits, so it may be worth checking out the disk usage for Sling in an hour or so. Thanks, Robert [1]: https://ci-builds.apache.org/job/Sling/job/modules/computation/console
