0
Monitoring is a critical part of any computer system that has been brought in to a production-ready state. No IT system exists in true isolation, and even the simplest systems interact in interesting ways with the systems “surrounding” them. Since compute time, memory, and long-term storage are all finite, it’s necessary at the very least to understand how these things are being allocated.
Why Does the Calico eBPF Data Plane Support Metrics?
Perhaps this question seems contrived. However, it’s always worth spending a moment thinking about reasons before adding any technical complexity to a distributed system! After all, they are already quite complicated! So why does the Calico eBPF data plane support metrics through Prometheus and Grafana?
Well, the Calico eBPF data plane is production ready and widely deployed, so a well-configured Kubernetes cluster with the Calico eBPF data plane correctly enabled will be stable and reliable. However, distributed systems are inherently complex and when dealing with them, it is generally good practice to instrument and baseline metrics wherever they are available. Doing so provides many benefits, especially for capacity planning, change management, and as an early-warning or smoke-testing system.
Additionally, seeing a running distributed system fully instrumented can be Continue reading