Back to all scenarios
Scenario #416
Scaling & Load
Kubernetes v1.23, on-prem

Kubernetes Downscaled During Rolling Update

Pods were prematurely scaled down during rolling deployment.

Find this helpful?
What Happened

Rolling update caused a drop in available replicas, triggering autoscaler.

Diagnosis Steps
  • 1Observed spike in 5xx errors during update.
  • 2HPA decreased replica count despite live traffic.
Root Cause

Deployment strategy interfered with autoscaling logic.

Fix/Workaround
• Tuned maxUnavailable and minReadySeconds.
• Added load-based HPA stabilization window.
Lessons Learned

HPA must be aligned with rolling deployment behavior.

How to Avoid
  • 1Use behavior.scaleDown.stabilizationWindowSeconds.
  • 2Monitor scaling decisions during rollouts.
Previous ScenarioNext Scenario