When I was working on one of my Kubernetes tasks in one of my projects, the added new annotations (key-value pairs) to my Ingress, brought down one of my applications. This gives me the result of learning one more concept, PDB (PodDisruptionBudget). This talk is a story of that failure — how and what is PDB
in K8s, misconfigured PDB, and unexpected node drain led to downtime, how I debugged the issue, and the practical steps I used to fix it. Along the way, I’ll explain:
- How Kubernetes handles evictions and node drains,
- Why PDBs sometimes make disruptions worse if not designed carefully, and the recovery strategies (scaling, uncordoning, adjusting budgets, curating nodes).
- How the node lifecycle works together