Jobs randomly hanging at "Starting your workflow run..."

Hello!

We’re trying to migrate our CI to Actions and we’re experiencing stability issues regarding jobs randomly being stuck at Starting your workflow run..., for example here https://github.com/HaxeFoundation/haxe/runs/905701602?check_suite_focus=true

This push is still rather new, but if we look through history, we can see jobs being stuck for days and we don’t understand if it’s something we’re doing wrong or there are some issues still with GitHub Actions itself.

Any help is appreciated!

2 Likes

Hi, it might be the case that you are hitting our concurrency limit:

https://docs.github.com/en/actions/getting-started-with-github-actions/about-github-actions#usage-limits

Please keep in mind this limit is for the entire organization.

Hi! We’re getting the same thing in the rubygems organization since a couple of days ago. Some jobs are queued but they end up never being started.

We haven’t done any changes recently that we would expect to increase our usage, so it would seem weird that we would be reaching some usage limits that we were not reaching before.

Thanks in advance for any help!

EDIT: Example of stuck workflow runs in this PR in case it helps: https://github.com/rubygems/rubygems/pull/3848.

I don’t quite understand how do these limits work. To my understanding, if we hit the limit, the future jobs should wait for the previous ones to complete. And anyway, we also trying to cancel previous runs when a new one starts, but that doesn’t seem to do anything with these stuck jobs. What am I missing?

Here’s another job triggered by a push 2 days ago: https://github.com/HaxeFoundation/haxe/runs/898181262 :man_shrugging:

@nadako @deivid-rodriguez

I’m very sorry – it might not be a concurrency issue. I followed up with my colleagues from the Actions product team and this might be an issue on our end.

We have an internal issue tracking this, and I’ve added both your examples to the list and it will be useful for the engineers to investigate.

Although we cannot guarantee an ETA on when this will be fixed, we can assure you it is being looked into. Thank you for your understanding!

6 Likes

We’ve run into this as well. The problem is exacerbated by not having the ability to re-trigger a single job, you have to re-do the whole run when these issues happen.

A couple more examples for you:


1 Like

This also started happening to me yesterday. It’s only affecting the job that uses the jobs.needs syntax.

Here’s a link to the failure. I’ve tried restarting several times to no avail.

Things seem to be working fine now. I’m thinking it might have been fixed already.

I have looked at the workflow runs that previous posters have submitted, and it appears that they have run into the problem I detail in Workflow run has completed successfully, but one job displays as "queued" in the UI : the task runner part of GitHub Actions runs all the jobs, but the UI part of GitHub Actions occasionally “misses” the start of one of the jobs. This results in one of the jobs permanently shows as “pending” in the UI - even when the entire workflow run shows as completed.