Back to all scenarios
Scenario #410
Scaling & Load
Kubernetes v1.22, Azure AKS
Application Didn’t Handle Scale-In Gracefully
App lost in-flight requests during scale-down, causing 5xx spikes.
Find this helpful?
What Happened
Pods were terminated abruptly during autoscaling down, mid-request.
Diagnosis Steps
- 1Observed 502/504 errors in logs during scale-in events.
- 2No termination hooks present.
Root Cause
No preStop hooks or graceful shutdown handling in the app.
Fix/Workaround
• Implemented preStop hook with delay.
• Added graceful shutdown in app logic.Lessons Learned
Scale-in should be as graceful as scale-out.
How to Avoid
- 1Always include termination handling in apps.
- 2Use terminationGracePeriodSeconds wisely.