Updating a K8s Cluster

Tips for how to make your applications resilient to Kubernetes cluster updates

1. Make sure you use readiness checks
2. Handle SIGTERM in your applications so they can gracefully shut down. Use a terminationGracePeriodSeconds configuration to give your application time to shut down.
3. When nodes are deleted/updated, this will cause the pods to be rescheduled. The edge case here is when all replicas for an application or on a single node – to avoid tihs you can use Pod Disruption Budgets (PDBs) to ensure that at least one replica is always available. Alternatively, you can use minAvailable to ensure that at least one replica is always available (recommended).