Deep Dive: Kubernetes Metric APIs using Prometheus
summary
Kubernetes now exposes three concrete metric APIs — resource, custom, external metric APIs — and a Prometheus adapter converts API requests into Prometheus queries using a Prometheus adapter configmap with resource_rules, series discovery queries, label→resource associations, regex-based naming and template variables. Example: HTTP requests total is transformed into HTTP requests per second → HPA via sum(rate(...)) and drives HorizontalPodAutoscaler scaling. Works in production, but config is complex, metric-server must be disabled and external metrics support is newly merged.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:44Problems with Heapster
- 3:16Meet the Metrics APIS
- 3:57Implementations
- 8:56Discovering Metrics APIs
- 15:04Custom Metrics API Service
- 16:04Conversion Kubernetes Prometheus
- 22:06External Metrics API Service
- 23:14Prometheus Adapter - Future
related talks
34mAutoscale your Kubernetes Workload with Prometheus - Frederic Branczyk, CoreOS
34mScale your microservices by MicroProfile Metrics on Kubernetes
25mAutoscaling Kubernetes Workloads Based On Metrics
37mMonitoring Kubernetes with Prometheus
30mHorizontal Pod Autoscaler Reloaded - Scale on Custom Metrics
31mEvolution of Metric Monitoring and Alerting: Upgrade Your Prometheus Today
19mKeynote: Monitoring Kubernetes Clusters with Prometheus
51mHorizontal Autoscaling with Kubernetes