GitLab

Created by: gtmtech

Gitlab v5.2.0 Gitlab server git version: 1.8.2.3 Gitlab client git version: 1.7

Rather than use the repo name provided by gitlab, a dev made up a URL for themselves using the HTTPS repo location, adding his name https://name@gitlab.some.domain/repo.git

It seems he was able to clone using this method, but when pushing, he took down the whole of gitlab. The POST request to /somerepo.git/git-receive-pack produced a 200, followed by a HTTP Status 502 2 minutes later (almost like a timeout). Every single subsequent request was a 502 to anything.

To rescue the situation I had to kill -9 puma, as it seemed the puma process had completely locked up.

Whilst this is a very weird set of events, I think gitlab shouldn't be locking up completely. Unfortunately there was nothing found in any logs (production.log, redis.log, puma.log, mysqld.log) under debug modes of any use in diagnosing what was happening.

I've no idea if the git versions were involved... After investigating I told the dev to use the git@gitlab SSH location format which solved their problem. However this issue is about gitlab locking up if somebody does try that.