Jobs on macOS-latest sometimes gets cancelled (An error occurred while provisioning resource)

[britzl]

We are from time to time seeing jobs running on macOS being cancelled.

The log for the job shows:

2020-01-23T11:48:23.5315674Z ##[section]Starting: Request a runner to run this job
2020-01-23T11:48:23.6273708Z Requesting a hosted runner in current repository's account/organization with labels: 'macOS-latest', require runner match: True
2020-01-23T11:48:23.6723606Z Labels matched hosted runners has been found, waiting for one of them get assigned for this job.
2020-01-23T11:48:23.6973639Z ##[section]Finishing: Request a runner to run this job

And the job annotations:

“An error occurred while provisioning resources (Error Type: Disconnect).”

“Received request to deprovision: The request was cancelled by the remote provider.”

This is really annoying as it is the last step of a fairly long chain of builds. Any idea why it happens and how we can avoid it?

[yanjingzhu]

Do you use a private repo or a public repo? If you use a public one, could you please share your repo links here? If you use a private one, I would recommend you contact GitHub Support or GitHub Premium Support. 

[britzl]

It’s a private repo. I’ll try to contact support. I thought GitHub representatives also responded to questions in the Community Forum…

[yanjingzhu]

According to this document, contacting support is a better way for your scenario. Thank you for your understanding.  

[xkubov]

Hi. I am having the same issue. Were you able to solve it with support?

[smatt989]

+1 having the same issue!

[xkubov]

Okay, so my problem was caused by an error in one of the steps - a missing dependency that would normally result in an error message. The thing is that no error message was provided after the cancelation of the job and therefore the debugging of the issue was really painful. 

[jcash]

No, I’m afraid we never got any answer from the Github Support.
The issue is still lingering, and failing many many builds for no good reason.

[yanjingzhu]

@jcash Sorry for the bad experience . There was an existing opened issue in our internal channel for this problem. Engineering team are still investigating on it.  I’m afraid they may still need sometime. 

[jcash]

That’s good to hear, I guess.

Have they found _anything_ we can go on?
We’re currently blocked by this.

[blane1988]

We are experiencing same error/warning at the end of a long build process with fastlane (gradle under the hood). Builds were working fine in the morning, and then failures began across all our app versions. So, either this is to do with CI environment or some dependencies.

With this issue, the failed steps don’t even produce logs, so we can’t debug.

[britzl]

Yeah, it is super frustrating. It is not happening very often for us, but when it does and it fails at the end of a long series of build steps it makes me want to look for an alternative.

[emmahodcroft]

Any update on this from anyone? We have run into this for the past few days and it’s incredibly frustrating. As others have said, there’s no logs, and so no way to debug. We get exactly the same error messages as shown in the above screen-shot.
Our repo is public (though not all keys are), so can share (if it helps anyone…): https://github.com/nextstrain/ncov-ingest/actions/runs/222496749

[jvesely]

In our case the failures appear to have been related to resource usage.
The feature page says each machine gets 2 virtual cores and 7GB RAM.
however, pytest identified 4 cores and ran 4 testing jobs in parallel. After it was artificially restricted to 2 jobs, the random ‘disconnect’/‘request to deprovision’ errors disappeared (and the jobs also run 2x slower than they used to)

[varunagrawal]

We’re currently encountering this issue with both MacOS and Ubuntu 18.04. Logs don’t provide any helpful info.

Any updates on this please?