Not exiting when runner creation errors out about ec2-github-runner

[hajapy]

Perhaps missing a clearInterval(interval) here: https://github.com/machulav/ec2-github-runner/blob/main/src/gh.js#L65

[hajapy]

Thanks for the fix. In my case I'm fairly confident the instance was able to connect to the internet (I've had many successful runners). I believe the polling rate is too high. At 60 api calls/minute, not accounting for any other api calls, we can only have about 16 minutes of waiting per hour per repo before reaching the rate limit.

I was thinking to drop it to poll every 10s and/or add a no-polling period for the first minute where it is very unlikely the runner will exist yet. This would help avoid unexpected issues for an active repo.

[machulav]

Perhaps missing a clearInterval(interval)

Good catch! 👍
Thanks!

It did not ever succeed to register, so the limit must have been hit before it could try to register.
Also, it seems likely that the interval used to check whether the runner is created should be increased or made configurable.

Based on my tests, even the smallest and least powerful EC2 instance type t2.nano is created and registered within 3 minutes.

If the runner is not registered in 3 minutes then, then most probably it won't be created because of the issue with the connection to the Internet or, as you said, in limitation of the GitHub API. However, I didn't face such a problem with GitHub API limits before.

That's why I would even reduce the timeout from 10 minutes to 5 minutes to decrease the billable time when the EC2 instance registration is failed.

[machulav]

@hajapy It's actually a good point! I will try to measure the average time of the self-hosted runner registration and then add a timeout before start pulling every 10 seconds.