heptapod_docker wrapper: fixed startup race condition
The `hg update` of `gitlab-rails` now being done in the Docker image rather than at startup, we have a race condition: `gitlab-ctl restart` gets executed before runsv is actually ready. This is frightening in all cases, and even leads to hard failure of the container on simple restarts. `gitlab-ctl reconfigure` is actually meant to start all services at container startup, and seems to be able to wait on runsv. This is the way it works in the GitLab CE image (as the comments in its `wrapper` script indicate. Apparently, we simply don't need the `gitlab-ctl restart` in the heptapod setup anymore. For safety, we also move the part that ensures that the hgserve service and its configuration files are where expected before the call to `gitlab-ctl reconfigure`.
Loading
Please register or sign in to comment