Skip to content

Increase replicas and remove liveness probe

Tan Le requested to merge 100-increase-replicas-and-disable-liveness-probe into main

What does this MR do and why?

When the model gateway is on high load, all web processes are busy serving external requests. The liveness probe is amongst those requests which got timed out while waiting for a response. K8s monitor will pick up these failures as the pod is unhealthy and trying to restart it. This results in the pod being in a CrashLoopBackOff state.

Relates to gitlab-com/gl-infra/production#14451 (closed) and gitlab-com/gl-infra/production#14455 (closed)

Edited by Tan Le

Merge request reports