Kubernetes makes monitoring harder in a specific way: the things you care about — a Deployment, a service — are made of pods that come and go. A tool that tracks pods tells you a lot about the last ten minutes and little about the last week. This guide is for small teams running a cluster or two without a dedicated platform team, and it's ordered from the simplest option to the most complete.
A disclosure: we build Archynt and it's on the list. Prices are left out on purpose — they change too often; check each vendor's page.
What to monitor in a cluster
Nodes: CPU, memory and disk pressure.
Pods: restarts, OOM kills, CrashLoopBackOff, pods stuck in Pending.
Workloads: Deployments and StatefulSets with fewer ready replicas than desired, HPA behaviour, failed Jobs.
Services: request rate, error rate and latency, and which services depend on which.
Logs and events: container logs and Kubernetes events, searchable after the pod is gone.
At a glance
Tool Type Best for
kubectl top + metrics-server…