Default gitlab-shell timeouts may be too short
ZD: https://gitlab.zendesk.com/agent/tickets/82309
Customer has a few large repositories that got killed due to the default 800 s timeout in git fetch
passed by GitLab to gitlab-shell, resulting in lots of temporary pack files that for some reason never got cleaned up. We bumped up the timeout via:
gitlab_rails['gitlab_shell_git_timeout']
This seems to have cut down the errors dramatically.
We should consider bumping up this default to a higher level for Geo--and possibly mirrors too.
Proposal
- increase default timeout
- add detailed documentation about time out in Geo troubleshooting docs and alert support
/cc: @nick.thomas, @tiagonbotelho
Edited by Nick Thomas