Monitoring KEDA

Another important aspect for operating KEDA is monitoring. This effectively means observing metrics that define the status or health of a system. So, in this case, we’re interested in observing metrics from KEDA’s components. In Chapter 3, you already used Prometheus and Grafana to monitor the resource utilization of a workload with the purpose of right-sizing the requests of CPU and memory. To monitor KEDA’s components, we’re going to continue using these tools and take advantage of what KEDA’s components already provide.

KEDA’s components already expose a certain number of metrics for Prometheus to scrape. As most of the work in KEDA is done in the operator, you’ll find that most of the metrics come from this component. However, KEDA also exposes metrics for KEDA’s metric server and KEDA’s admission webhook. All these metrics can be scraped by Prometheus on port 8080 at the /metrics endpoint.