Metrics and Profiling for Kubenetes microservices [closed]
The best open source monitoring stack for Kubernetes is the one based on Prometheus (for metrics collection and storage) and Grafana for dashboards and visualisations.
Here is an article which describes how it works, there are plenty of them on the web.
In a nutshell, this is how it works:
- Prometheus and Grafana are independent deployments in your cluster (potentially in a dedicated
-infranamespace), configured via ConfigMaps. - Prometheus needs to be configured to scrape metrics from the pods on the application's namespace
- The application needs to expose metrics on a specific HTTP port, and path, e.g.
/metrics - Grafana is by default connected to Prometheus, so no other config has to be done on that side (maybe just double check the datasource config)
Almost nothing needs a deployment into the control plane, it just needs to run somewhere. But that could be in your namespace or anywhere else. Elastic has a FOSS APM product that I've heard decent things about, or you can look at manually instrumenting stuff for Prometheus.