Runner health check
GitLab CE version 8.16.5
Every few days I have problem with a runner. It is always related to some Docker issue. The problem is, it quickly pollutes all my pipeline results, because the pipelines don't differentiate between "broken build" and "broken runner". Having to manually retry all broken pipelines is tedious.
I think there is a big need of some automated runner health check system. Some of the issues could be detected on build level - runner starts by pulling an image from registry, and if that results in "Bad response from Docker engine" (or a similar error), the runner should be automatically disabled. For other errors, some kind of "never allowed to fail" job could be defined (eg. run hello-world
image and compare output).