I noticed that for the last 24 hours or so the Linux build jobs have been capped at just under 600 from 13:30 on Thursday (whereas it has peaked over the last week in the 1200 range) which is a change from previously. This is with no announcement and NO reports of any incidents. As a result the backlog is ramping up and is in excess of 6,000 (it reached a local minimum of ~900 in the middle of the Thursdays-Friday night whereas it has always cleared in the past)!
That isn’t immediately clear as an explanation - as far as I can tell the repositories for the project I code from have already made the switch - it is more as if they have said, "oh we do not have to provision quite as many build workers for the free projects, let’s take off a third and see if they squeal, if not we can save some Euros or use them elsewhere… "
"oh we do not have to provision quite as many build workers for the free projects, let’s take off a third and see if they squeal, if not we can save some Euros or use them elsewhere… "
Right. I hope Travis will fix it or get the workers back. I’ve been using it for last 3 years and it did a great job for my open source project. As of now, the project development is in collapse cause I have build and publishing process automated as well
Yeah, the cap has now dropped to around 560 jobs - which was only briefly hit on Saturday but we have been hard at that since 15:00 UTC on Sunday though it might just be descending from that peak now it is 23:00 UTC Sunday night.
Of course the situation is not going to improve tomorrow without action by the Travis people - and as things stand I wouldn’t be surprised if the backlog exceeds 10,000 (10K) by Wednesday and does not fall to less than 3K for the rest of the week-days after that point…
Yes, we are hitting this problem as well. It’s going to cause major problems very soon if it’s not resolved. Or, I suppose I can migrate over to travis-ci.com.
Will it really fix the issue? The ‘Travis CI Status’ link from from *.org and *.com leads me to the same page. If *.com projects builds on different pool of machines, how can I look for a status of the queue? Just to be sure that it’s reasonable to move to *.com now and that it will solve the issue we’re all facing.
The project I code for had made the switch some time ago and that is being impacted as far as I can tell.
The Linux builds maxed out again at 07:00 UTC and have been hard against that limit (558) ever since - the back log ramped up from zero at that time and plateaued today at around 8K at 15:45 UTC - and has not significantly dropped from that since then…
I’ll try this with a few of my repos soon—the ~535 worker cap is getting really annoying, as every weekday now, my builds basically don’t start until a day later!
And after a few minutes trying everything under the sun, I finally found there’s a ‘Sign up for the beta’ button on my travis-ci.org that I had to click to enroll my “organization” (which was actually my personal account, which already had many projects enrolled on travis-ci.com… so it was unintuitive, but now I can migrate repos.
Hope for the first one
