What will be included in the second iteration of a horizontally scaled GitLab? We have chosen to split components even more and change the shared filesystem type. We will introduce two monitoring components.
The following diagram is of the second iteration of the High Availability (HA) solution for GitLab:
The frontend servers in the basic HA design run the Unicorn/NGINX Rails application, as well as the background jobs, via the Sidekiq component. These are all installed as part of the GitLab omnibus package. In the second iteration, you can see that the NFS server in the diagram has been replaced with a Gitaly server. Gitaly is an Remote Procedure Call (RPC) server that is specifically designed to handle Git traffic and has the ability to scale better than NFS servers, while using less bandwidth.
In the basic HA solution, the shared filesystem...