GKE node upgrade stuck: PodDisruptionBudget blocks the drain

A cluster or node pool upgrade stalls because a PDB doesn't allow evicting the pods on the node.

Exact message you may see Cannot evict pod as it would violate the pod's disruption budgetnode pool upgrade stuck PodDisruptionBudgetGKE upgrade blocked by PDBEviction would violate the pod's disruption budgetPodDisruptionBudget drain
Seen on: Google Cloud Kubernetes

Meaning

GKE respects PDBs for up to an hour during upgrades. Relax overly strict PDBs (minAvailable: 100%, single replica with minAvailable: 1), add replicas, or temporarily patch the PDB. Check kubectl get pdb -A for ALLOWED DISRUPTIONS = 0.

Common causes

  • PDB with ALLOWED DISRUPTIONS 0
  • Single-replica deployment with minAvailable 1
  • Pods stuck terminating
  • Orphaned PDB selecting nothing
  • Too-strict maxUnavailable

⚡ Quick fix

  1. kubectl get pdb -A
  2. Increase replicas or loosen the PDB
  3. Delete/patch the blocking PDB temporarily

Detailed fix by platform

Shell

  1. bash
    kubectl get pdb -A
    kubectl get pods -A -o wide | grep NODE_NAME

How to diagnose

  1. Allowed disruptions — 0?
  2. Replicas — More than 1?
  3. Pods — Terminating?

🔧 Still not fixed?

Many errors look alike. If the steps above didn’t solve it, one of these is probably what you’re facing:

Where this error sits

Origin: Platform & cloud. Errors that are often confused with or caused by this one (click to open):

Autopilot — GKE Autopilot rejected the workload: resources adjusted / violates Autopilot constraints (Google Cloud / Firebase)AutopilotFailedScheduling — Kubernetes: Pod stuck in Pending — FailedScheduling: 0/3 nodes are available (Kubernetes)FailedSchedulingGKE Ingress: backends marked UNHEALTHY (502) — GKE Ingress: Some backend services are in UNHEALTHY state / 502 Server Error from the load balancer (Google Cloud / Firebase)GKE Ingress: backends…gke-gcloud-auth-plugin not found — kubectl: error: executable gke-gcloud-auth-plugin not found — it looks like you are trying to use a client-go credential plugin that is not installed (Google Cloud / Firebase)gke-gcloud-auth-plugi…GKE: 0/3 nodes are available: Insufficient cpu — GKE/Kubernetes: Pod unschedulable — 0/3 nodes are available: 3 Insufficient cpu (Google Cloud / Firebase)GKE: 0/3 nodes are av…GKE: Quota 'CPUS' exceeded — GKE/Compute: Quota 'CPUS' exceeded. Limit: 8.0 in region us-central1 (QUOTA_EXCEEDED) (Google Cloud / Firebase)GKE: Quota 'CPUS' exc…ImagePullBackOff — GKE ImagePullBackOff: 403 Forbidden pulling from Artifact Registry (node service account lacks reader) (Google Cloud / Firebase)ImagePullBackOffPDB

Explore the map of this error Full error map →

🧠 Still stuck? Analyze your error

Paste the full message, response headers or stack trace — we'll detect the platform and point to the most likely cause.

💬 What worked for others

No tips yet — did something else fix this for you?

Share what worked for you

Tips are checked by an editor before they appear. Links are removed.