1. Proposed title of this feature request
Manual setting or auto-tune of Requests for ovn pods in ovn-operator
2. What is the nature and description of the request?
Increasing stability of cluster / stability of network traffic om highly utilised clusters The ovn operator should either auto-tune requests by using metrics / VPA so that requests matches the average of the running ovn-pods. On a highly utilised cluster we have an example: Average is 200 mcores for the individual ovn-pod, while default requests is 80 mcores. This means that when the node is extremely well-used then the ovn-pod is starved of 120 mcores - thus introducing network issues. Network issues increases load further - so it's a vicious circle.
Allowing either "auto-tuning" of requests by looking at a) cluster-size or b) using Metrics or c) using VPA would be good. Or d) at the very minimum give the consumer the ability to set the requests manually, if they have the know-how.
3. Why does the customer need this? (List the business requirements here)
Stability! Network stability! Extremely used nodes can lose network traffic or get slow network traffic with increased latencies etc.
4. List any affected packages or components.
OVN operator ovn-kubernetes pods.