Hey guys, I’m currently learning how to manage Kubernetes infrastructure and today’s issue is with the autoscale stuff. I created my cluster on Digital Ocean an installed some of the one click add-ons, such as prometheus monitoring.
Querying all the pods of the cluster I noticed that only 3 were mine and the other 10 - 15 ones were due to the add-ons installed. Initially, this didn’t represent an issue, but after messing with the HPA I thought that these pods may have been using cpu and memory resources that would be better suited for the pods of my application.
Is there any problem in restricting all the metrics pods to one specific node to better organize the app? So if a node has 1vCPU, I can assing e.g. 25% for each pod without having to compensate for metrics pods.
I’m happy to hear other ideas and tips as well, thanks!