travis-ci no longer running. #17824
Comments
|
Indeed, the ppc, arm64, and s390x builds still left on travis stopped two days ago. When I login and go to the build page, I see "Builds have been temporarily disabled for public repositories due to a negative credit balance. Please go to the Plan page (which is empty) to replenish your credit balance or alter your Consume paid credits for OSS (which is also empty) setting." I admit I have little desire to deal with this: IBM and travis should work out an exclusion for NumFocus (at least) projects if IBM wishes us to continue maintaining software for their hardware. We still have a job running on shippable for arm64. |
|
Note we should really verify we support Apple silicon by running a native build there as well. |
|
Until now, none of the NumFOCUS projects seem to have any reasonable experience with keeping those things running. If we need very few resources for something, it may be possible. But aside from that, we probably have to assume that Travis is a lost cause right now. |
|
Yeah, SciPy and MDAnalysis are both facing the same problem. For SciPy, it is a double hit because the wheels repo still has a massive Travis matrix as well. A few people were talking about chipping in to hold things over for a bit, but when I look at the options it is really bad. Even if I'm like "ok, I'll shell out $70 in the main repo and $70 in the wheels repo to get SciPy 1.6.0 out", it would still be at least 5x slower than the previous 5-concurrent free plan.
I emailed Leah and Andy at NumFOCUS; Andy said they might be able to work something out with CircleCI or Azure, but I've heard no indication that NumFOCUS projects have propsects with Travis CI arragements at this time. I emailed Travis CI support on Sunday and have not heard anything back.. |
|
NumPy needs something to build arm64 wheels on, apart from that I think we can get by without travis. At some point I expect azure to support arm64, not least for Windows on arm64. $279 doesn't seem too bad for a month. i was thinking of chipping in ~$1000 if it could be targeted. Or go for a slower build and just spend time. We could all use more sleep :) |
|
Hello everyone, |
|
@ayush1300 you are commenting on an issue about using travis ci. You can reach out on the mailing list or come join one of our community meetings. @tylerjereddy would you like help migrating scipy to github actions/azure? As for arm64: NumPy can get away with using shippable instead of travis, but I think the scipy build times out, it needs the more powerful graviton machines only available on travis. Edit: fix the link |
|
Can someone from the NumPy twitter account ping again e.g., https://twitter.com/neoscms/status/1330829574054273025 They seem to be active. |
|
@ilayn it might be more effective for NumFocus to reach out to support@travis-ci.com and negotiate for all their sponsored projects. |
|
Yes also true |
|
Travis CI replied to IBM that they provide free credits to Open Source projects that request more credits. Has anyone actually asked Travis CI? Not even the "sponsored project" designation. Just ask for more credits. |
|
The statement from Travis CI was:
I don't know if one needs to request more credits incrementally or one needs to request a larger block of credits once one has determined the monthly need. I don't know that the Travis CI support team investigates in detail to estimate the necessary amount. An initial request of additional credits is an initial request. I'm not suggesting that Travis CI is making this easy, but the response seems like an overly quick declaration that no additional credits are available and the process is unworkable. |
astropy might indeed relied more on travis than other projects, mainly due to the ecosystem nature of the project, it wasn't just the core package building on travis but a lot of other repos in the organisaion. We have already evaluated transitions to pipelines or actions previously, but not all the features we used were available, and anyway, we were operating from the POV of not fixing things that are not broken (travis was always a very well understood CI provider which many of our contributors were familiar with, and we haven't really run into queue bottlenecks in the past 1-2 years). Anyway, at the beginning of the month, when the announcement came out, we put together a plan to migrate and keep only the unique stuff on travis, while we were also exploring the possibilities of getting credit or paying for the services. We asked for a monthly credit allowance based on our past usage, and indeed got a fraction of it, as a one of credit, that promply run out in a few days even though some repos has already been migrated away. Some numbers are here: astropy/astropy-project#164 |
|
Oh, and to the opening comment:
I believe it's an overall, rather than a monthly quota. |
|
Hello folks, please note that I have added additional credits to this organization as directed. Thank you cc: @pauladamgordon |
|
Hi @muizzk thanks! How can we see the current state of our account? |
Looks like travis.ci has run out it's monthly quota.
The text was updated successfully, but these errors were encountered: