Back to all scenarios
Scenario #409
Scaling & Load
Kubernetes v1.25, Prometheus Adapter

Custom Metrics Adapter Crashed, Breaking Custom HPA

Custom HPA didn’t function after metrics adapter pod crashed silently.

Find this helpful?
What Happened

HPA relying on Prometheus metrics didn't scale for hours.

Diagnosis Steps
  • 1kubectl get hpa: metric unavailable.
  • 2Checked prometheus-adapter logs: crashloop backoff.
Root Cause

Misconfigured rules in adapter config caused panic.

Fix/Workaround
• Fixed Prometheus query in adapter configmap.
Lessons Learned

Custom HPA is fragile to adapter errors.

How to Avoid
  • 1Set alerts on prometheus-adapter health.
  • 2Validate custom queries before deploy.
Previous ScenarioNext Scenario