3 ms·
Out of curiosity, why do you recommend running an Omnibus container instead of separate containers for Gitlab, Postgres and Redis? It seems like it only gains
by teraflop 12y ago
Out of curiosity, why do you recommend running an Omnibus container instead of separate containers for Gitlab, Postgres and Redis?
It seems like it only gains a tiny bit of convenience (one "docker run" command instead of three), and you lose a lot of flexibility to move things around and reconfigure later on. What if I want to run multiple frontend instances, or keep a warm standby database replica?
- sytse 12y agoThe main reason is that the Omnibus package setup has additional logic to respawn workers that are using too much memory. I also like that it uses a separate data container to store everything.
- codereflection 12y agoIn the https://github.com/sameersbn/docker-gitlab https://github.com/sameersbn/docker-gitlab container, you can mount volumes from the Docker host to store data as well, and it also uses linked containers for Redis and Postgres. The respawn workers logic does seem important though. Does the manual gitlab install also handle situation?
- sytse 12y agoNo, the manual install doesn't handle that. It requires Runit if I remember correctly.