Back to all scenarios
Scenario #415
Scaling & Load
Kubernetes v1.25, custom metrics + Prometheus Adapter

Prometheus Scraper Bottlenecked Custom HPA Metrics

Delays in Prometheus scraping caused lag in HPA reactions.

Find this helpful?
What Happened

HPA lagged 1–2 minutes behind actual load spike.

Diagnosis Steps
  • 1prometheus-adapter logs showed stale data timestamps.
  • 2HPA scale-up occurred after delay.
Root Cause

Scrape interval was 60s, making HPA respond too slowly.

Fix/Workaround
• Reduced scrape interval for critical metrics.
Lessons Learned

Scrape intervals affect autoscaler agility.

How to Avoid
  • 1Match Prometheus scrape intervals with HPA polling needs.
  • 2Use rate() or avg_over_time() to smooth metrics.
Previous ScenarioNext Scenario