What is the webserver of Gitlab when deployed into Kubernetes

General
1

Hi there,
we have a Gitlab setup on Kubernetes and we are trying to figure out, which component forms the Webserver for Gitlab - We are using the reference architecture for 3000 User/60 RPS.

When deployed on a regular server it should be Puma, but it seems that its different when its deployed on Kubernetes.

I hope my request is clear so far, let me know if there is something else you need to know.

2

Hi there,

Yes, despite the different packaging approach on Kubernetes, Puma (plus GitLab Workhorse) is still what handles the web/Rails traffic for GitLab from my understanding.

  • In Omnibus installations, NGINX proxies requests to Workhorse, which in turn dispatches them to Puma (the Rails application server).

  • In the official GitLab Helm chart, things are split out into separate pods:

    • A webservice pod (the main Rails application) – which includes Puma and GitLab Workhorse in a single container image.
    • An Ingress Controller or Load Balancer (generic Kubernetes tooling) typically handles SSL termination and then forwards traffic to the webservice pod.

In other words, on Kubernetes it’s still Puma + Workhorse, just deployed differently (separate pods instead of the all-in-one Omnibus approach). The reference architectures do not swap in another webserver; they just bundle these same components under the “GitLab webservice” umbrella.

Cheers

1 Like