Pod Disruption Budgets and Voluntary Disruptions

You run kubectl drain during a maintenance window. The command hangs for ten minutes, then you cancel it and try again with --force. Five minutes later, your monitoring stack goes dark because every Prometheus replica was evicted at once. The node is drained. Your observability is gone.

PodDisruptionBudgets exist to prevent exactly this. A PDB is a contract between you (the application owner) and the cluster operator: it declares how many pods your application can afford to lose during voluntary disruptions, and the Kubernetes eviction API enforces that limit automatically. In the CKA exam, you will create PDBs, drain nodes that have them, and debug the cases where they block evictions.