Investigate and Set Resource Limits for ArgoCD to Prevent Reconciliation Loop-Induced Node Overload

[sedatbasar]

If ArgoCD gets stuck in a reconciliation loop, it can push up the load on a node to the point where the node becomes unusable. When Kubernetes reschedules the ArgoCD controller onto other nodes because one node becomes "NotReady", it can make the whole cluster unstable. It's highly recommended to observe the baseline memory and CPU requirements for ArgoCD and set resource limits accordingly.
I estimate it would take 1 hour to study ArgoCD's baseline resource usage and another 2 hours to test and apply the settings. I would ask that everybody in the DevOps team making these changes report back on what the baseline ArgoCD resource usage is and what limits they set. This will be valuable knowledge to share amongst the team.