On Tue, 24 Nov 2020 at 14:44, Ganesh Murthy <gmur...@redhat.com> wrote: > > On Mon, Nov 23, 2020 at 11:12 AM Robbie Gemmell <robbie.gemm...@gmail.com> > wrote: > > > Hi folks, > > > > Short version: > > 1) The Travis build jobs will migrate to a new URL, on their .com site > > rather than .org, links etc will need to be updated. > > 2) Travis have introduced relatively tiny default resource limits for > > free users, so you may want to disable Travis on your Github forks to > > save them until really needed (or you can apparently individually > > apply for a higher limit). > > > > Expanded version: > > First up, as some of you may know Travis CI is migrating away from use > > of https://travis-ci.org to solely using their other > > https://travis-ci.com site. This has been underway for many years at > > this point with no real traction, but a final deadline of Dec 31st has > > been set for the .org bits to become defunct, and worker nodes have > > been migrating across for weeks now, so the point has come a switch is > > required. > > > > When done, the existing URLs will just give a landing page saying the > > build moved, with a link to go to it. Folks using the existing URLs > > for build status etc will need to update their references. > > > > Infra started to migrate some jobs over themselves, and also migrate > > the paid ASF concurrency limits plan across for the apache github org. > > They decided they didnt want to end up migrating >2000 jobs when many > > aren't really used anymore, and so have asked for the remainder that > > migrations be requested. > > > > None of ours appear to have been moved yet, so I have requested [1] > > that infra do the migration for these repositories: > > qpid-broker-j > > qpid-dispatch > > qpid-jms > > qpid-proton > > qpid-proton-j > > qpid-site > > Thanks for doing this! So it looks like no other additional action is > required on our side. > Once this migration is complete, we will be as usual able to access the > Travis > builds from the github <https://github.com/apache/qpid-dispatch> website > with the just the one difference that the Travis build URLs > will point to the .com website. >
Should be the case yep. > With regards to the resource limits on forks, it looks like we will have to > wait to see if they automatically > enable builds or if they will make us enable it. If they don't enable it, > I don't plan on enabling it. > Again, no action required from our side if they don't enable it. > Yep. > Qpid Dispatch has a mac build and it looks like we might have to pay for it > if we want to keep it around by purchasing what they > call credits (1 minute of mac build time costs 50 credits). Do we want to > do this? Or are we just going to drop the mac os builds? > We would surely drop it from the Travis build in that case. GitHub Actions also has MacOS build nodes, so that build workflow could be updated to use them if desired. > > > > Secondly, note that Travis have significantly changed the terms for > > 'free' usage on travis-ci.com during this process, to 'combat abuse'. > > The effect is that it severely curtails usage for non paying folks, or > > folks who dont fill out a form to get an > > individually-requested/tailored 'allotment of OSS minutes': > > https://blog.travis-ci.com/2020-11-02-travis-ci-new-billing > > > > I know the ASF have paid for additional resources at Travis for a long > > time, so I dont think this really impacts things for the main > > repositories, but I think this probably presents an issue for > > developers forks. As such, folks might want to disable the Travis > > integration on their forks to stop it burning the very limited .com > > resource you will have available by default. Folks working on > > components that dont have Github Actions / Appveyor / Jenkins / Other > > builds in addition to Travis, might also wish to establish some as an > > alternative for PRs etc. > > > I did not realize that the ASF folks were paying Travis (for additional > resources). I always thought > that Travis was free for open source projects. > There are resource limits applied within a given github org, and since the ASF projects are all under the apache org that makes for a lot of contention for the free resources. Particularly with some projects having monstrous builds (in terms of length and/or executor usage) utterly drowning out smaller or less frequent builds. The basic limits just wouldn't work for the org, you'd probably be waiting days or more for builds to run, so ASF infra pays for increased concurrency such that the backlog more manageable (and even then, its still fairly typical jobs have to wait some during peak times, certainly longer than they do on a personal fork building outside the org limits...this may ease as certain larger projects decide to use other services instead though, as the use-shiny-new-CI-service loop repeats once again, with extra prodding from these changes). > > > > Robbie > > > > [1] > > https://lists.apache.org/thread.html/r2863a94f8a37986594f44bd36b16d4065e9d09c25c90fc3b5f052e41%40%3Cbuilds.apache.org%3E > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: dev-unsubscr...@qpid.apache.org > > For additional commands, e-mail: dev-h...@qpid.apache.org > > > > --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@qpid.apache.org For additional commands, e-mail: users-h...@qpid.apache.org
