VPA Not honoring maxAllowed Memory Limit

[kmsarabu]

I am encountering an issue with the Vertical Pod Autoscaler (VPA) where it does not honor the maxAllowed resource limits for memory. Below is the VPA definition I am using:

apiVersion: "autoscaling.k8s.io/v1"
kind: VerticalPodAutoscaler
metadata:
  name: test-vpa
spec:
  targetRef:
    apiVersion: "apps/v1"
    kind: Deployment
    name: test-deployment
  updatePolicy:
    updateMode: "Auto"
  resourcePolicy:
    containerPolicies:
      - containerName: '*'
        controlledResources: ["cpu", "memory"]
        minAllowed:
          cpu: 500m
          memory: 4Gi
        maxAllowed:
          cpu: 4
          memory: 16Gi

After running a CPU stress test, the resulting resource limits observed on the pods are:

Limits:
  cpu:     4
  memory:  32Gi   <- this is more than VPA Object's MaxAllowed->memory?
Requests:
  cpu:      500m
  memory:   4Gi

Despite setting the maxAllowed memory limit to 16Gi, the VPA scaled the memory up to 32Gi.

Steps to Reproduce:

1. Deploy a VPA with the provided configuration.
2. Apply a CPU stress test to the target deployment.
3. Observe the memory and CPU limits on the autoscaled pods.

Expected Behavior: The memory limit should not exceed the maxAllowed value of 16Gi.
Actual Behavior: The memory limit scales up to 32Gi, exceeding the maxAllowed value.

Could there be any known issues or configurations that might lead to this behavior? Thank you in advance for your help!

[adrianmoisey]

/area vertical-pod-autoscaler