Custom compute class creates a lot of duplicated node pools

[kkrampa]

Hey,

we started using Custom Compute Classes.

We created the following config

apiVersion: cloud.google.com/v1
kind: ComputeClass
metadata:
  name: cost-optimized
spec:
  priorities:
  - machineFamily: c2
    spot: true
    minCores: 32
  - machineFamily: c2
    spot: true
  - machineFamily: c2d
    spot: true
  - machineFamily: n2
    spot: true
  - machineFamily: n2d
    spot: true
  - machineFamily: c2
    spot: false
    minCores: 32
  activeMigration:
    optimizeRulePriority: true
  autoscalingPolicy:
    consolidationDelayMinutes: 5
    consolidationThreshold: 70
  nodePoolAutoCreation:
    enabled: true

 It created about 10 duplicated nap-c2-standard-60-spot node pools. Is this desired behaviour? I was expecting single node pool for c2-standard-60 spot. 

We're running k8s 1.30.5-gke.1014001

[shannduin]

Could you also provide a scrubbed manifest for the workloads that you deployed to use this class? 

[kkrampa]

Are you interested in anything in particular in these manifests? We deployed a lot of workloads using this class (~20 deployments which overall have 2000+ of pods)

[shannduin]

Ah yeah I was trying to see why it was creating multiple node pools based on your node selectors/affinity/resource requests. How many nodes are in each of those pools?

[kkrampa]

This is the current state

[shannduin]

Are Pods from different deployments landing in the same node pool's nodes, or is every node pool dedicated to a specific deployment's Pods? I'm trying to see if it's trying to separate different deployments

[kkrampa]

Different deployments are landing on the same node pool's nodes.