Questão de Service Configuration #019e3a4f

The correct tool combination is Prometheus and the prometheus-adapter. The HorizontalPodAutoscaler (HPA) in Kubernetes can scale pods based on custom metrics, not just CPU/memory. To use application-generated metrics, you need a metrics adapter that implements the custom.metrics.k8s.io API. Prometheus collects metrics from applications, and the prometheus-adapter (also known as k8s-prometheus-adapter) exposes those metrics to the Kubernetes API server, allowing the HPA to query them.

How it works:

1. Prometheus scrapes custom metrics from application endpoints (e.g., /metrics).
2. The prometheus-adapter is configured with rules to map Prometheus metrics to Kubernetes custom metrics.
3. The HPA references these custom metrics in its configuration (e.g., type: Object or type: Pods with a metric name).
4. The HPA queries the custom.metrics.k8s.io API, which the adapter serves, and scales pods accordingly.

Why the incorrect options are wrong:

• Graylog is a log management system, not a metrics collector. It does not provide a metrics API for HPA. The graylog-autoscaler metrics and kubernetes-adapter are not standard Kubernetes components.
• Grafana is a visualization tool; it does not expose metrics to the Kubernetes API. While Prometheus is correct, Grafana alone cannot serve custom metrics to HPA.

Only Prometheus with the prometheus-adapter enables HPA to use custom application metrics for scaling.