OpsCortex
Back to archiveSelf-Healing and Scaling

Kubernetes Self-Heals Because of One Loop, Not Magic

Rollouts, autoscaling, and graceful shutdown become one idea once you learn reconciliation.

Production idea

ReplicaSets, rollouts, HPA, VPA, maxSurge, maxUnavailable, and draining.

Command set
kubectl rollout status deploy/checkoutkubectl get hpa checkoutkubectl rollout undo deploy/checkout
Monday morning
  • Set requests before using HPA.
  • Use maxUnavailable 0 and maxSurge 1 for capacity preservation.
  • Add preStop and SIGTERM handling.
  • Avoid HPA and VPA fighting on the same signal.