I have been experiencing a recurring issue with PDBs during every rolling restart or deployment. The issue manifests as a momentary event where a terminating pod from the old ReplicaSet is treated as unmanaged, causing the following warning event:

Warning CalculateExpectedPodCountFailed 49s (x3 over 49s) controllermanager Failed to calculate the number of expected pods: found no controllers for pod "service-6958844f49-rw9cr"

Just before the pod is deleted, the PDB incorrectly classifies it as an unmanaged pod, leading to the above warning.
Despite this being a transient issue (lasting only momentarily before the pod is removed), I have observed it consistently across multiple microservices.

Expected Behaviour: The terminating pod should still be recognized as managed by the ReplicaSet and not cause a miscalculation in expected pod counts.

Environment Details:
Kubernetes Version: 1.30 EKS

PDB Configuration:

  apiVersion: policy/v1
  kind: PodDisruptionBudget
  metadata:
    labels:
      app.kubernetes.io/managed-by: Helm
      app.kubernetes.io/name: test-pdb
      argocd.argoproj.io/instance: test
      helm.sh/chart: generic-1.0.0
    name: test
    namespace: test
  spec:
    maxUnavailable: 1
    selector:
      matchLabels:
        app.kubernetes.io/name: test