A node drain or maintenance window keeps failing even though the operator can count enough replicas in the cluster.
PDB allows disruption on paper but drain still stalls because the controlling workload already has an unavailable revision elsewhere
Replica count and budget look safe locally, yet the workload already consumed its allowance on another node or rollout shard.
Scenario
What to check first
- Identify the primary failure signal in the Distributed Availability Budget scenario.
- Separate visible symptoms from the underlying technical dependency.
- Describe the safest recovery path and the follow-up prevention work.
Checking checklist
- Summarize the current impact and the last known change.
- Collect direct evidence from logs, runtime state, and configuration before changing anything.
- Separate immediate recovery from permanent prevention work.
Recovery and prevention
Look at rollout-wide unavailability before forcing eviction or removing the PDB.
Questions worth viewing together
Community-field Kubernetes maintenance problem inspired by Stack Overflow cases where PDB allowance was already consumed by another unavailable revision. Local replica math can be misleading if another revision already consumed the disruption budget.
Teams often assume drain is buggy when rollout churn elsewhere is the real blocker.
PDB decisions are workload-wide, not scoped to the node you happen to be draining.
Similar cases seen in the field