I have several runners (each capable of running multiple workers/jobs at once) across multiple machines. My pipeline has stages that contain multiple/parallel jobs. Sometimes, a runner will get “greedy” and try to run all of the offered load from a stage by itself (instead of letting an otherwise idle peer take a job).
Is there some way to meter the frequency that a runner checks for new jobs to run or otherwise control the rate that it bids-on/takes a ready job (in hopes of allowing an idle peer to successfully bid on said job)?
I was searching for the meaning of check_interval in /etc/gitlab_runner/config.toml when I came across
I still haven’t found out what that setting actually does, but I think that might come closer to a solution to your problem, but for now that issue might be a little interesting, even though I don’t see how it can help you.