[Review Apps] Add resource requests and limits to most of the chart resources
What does this MR do?
- I noticed that some VM instances (i.e. nodes) were at 100% CPU.
- At the same time, I noticed that the
gitlab-runner
deployment doesn't have resource requests. This is a problem because GCP autoscaling is based on the resource requests, not the actual resource usage (unless it hits the triggers set for the node, e.g.MemoryPressure
). - I also noticed that the
gitaly
andgitlab-shell
deployments were using more than their requested resources, which again could lead to non-optimal autoscaling decisions from GCP, leading to degraded Review App deployment timings, or even timeouts.
- I've decided to specify resources requests and limits for pretty much all the deployments/pods of the chart, based on actual numbers from the deployed Review Apps that we have.
- At the same time, I noticed that the
gitaly
gitlab-shell
sidekiq
unicorn
redis
minio
gitlab-runner
gitlab-monitor
postgresql
registry
Does this MR meet the acceptance criteria?
Conformity
Edited by Rémy Coutable