Scheduler job failures
While checking if runners behave correctly after upgrade, I noticed that some jobs are failing with message "The scheduler failed to assign job to the runner, please try again or contact system administrator".
Here are some hints:
- This is a server-side failure, having nothing to do with the runner itself
- There seems to be some automatic retry (to be confirmed)
- The prime suspect is workers saturation in HGitaly, as the CI/CD parts of the Web application are severely hammering the HGitaly server for changeset details.
Please, be patient about this. It will take some time to bring hgitaly#118 (closed) to a point where it can take care of the load generated by CI, but this is the proper way to solve this.