Tune a PodDisruptionBudget minAvailable

Set a disruption budget that protects availability during drains without blocking voluntary disruptions entirely.

~25 minkubernetesreliabilityops

Task

Configure a PodDisruptionBudget that keeps enough replicas during node drains.

When to use this

A node drain hangs because the disruption budget requires all replicas to stay available.

Prerequisites

A deployment with multiple replicas
kubectl access
An understanding of the availability target

Steps

1. Set a budget that leaves room to drain.

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: app-pdb
spec:
  minAvailable: 2
  selector:
    matchLabels:
      app: api

Expected output: At least two replicas stay available while one can be disrupted at a time.

2. Verify a simulated drain respects the budget.

kubectl drain node-1 --ignore-daemonsets --delete-emptydir-data --dry-run=server

Expected output: The drain proceeds because the budget allows one disruption.

Troubleshooting

Problem

Drain still blocked

Solution

Ensure replicas exceed minAvailable so there is room for one disruption.

Problem

Rollout stalls with maxUnavailable 0

Solution

Reconcile the PDB with the deployment strategy so both can proceed.

Limitations

PDB does not guarantee capacity, only disruption limits.
Single-replica workloads cannot tolerate any disruption.
Requires replicas above the budget.

What this recipe does

Configure a PodDisruptionBudget that keeps enough replicas during node drains.

A node drain hangs because the disruption budget requires all replicas to stay available.

Prerequisites

A deployment with multiple replicas
kubectl access
An understanding of the availability target

Procedure

Set a budget that leaves room to drain.

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: app-pdb
spec:
  minAvailable: 2
  selector:
    matchLabels:
      app: api

Expected result: At least two replicas stay available while one can be disrupted at a time.

Verify a simulated drain respects the budget.

kubectl drain node-1 --ignore-daemonsets --delete-emptydir-data --dry-run=server

Expected result: The drain proceeds because the budget allows one disruption.

How to confirm it worked

Run the commands above and confirm every expected result in order. If a step does not produce the stated output, use the troubleshooting notes before continuing. A passing run shows a node drain completes within the availability target while the budget keeps the required replicas serving.

Troubleshooting

Drain still blocked — Ensure replicas exceed minAvailable so there is room for one disruption.

Rollout stalls with maxUnavailable 0 — Reconcile the PDB with the deployment strategy so both can proceed.

Limitations

PDB does not guarantee capacity, only disruption limits.
Single-replica workloads cannot tolerate any disruption.
Requires replicas above the budget.