Connect your stack
A cluster in KubeMeridian is one datasource instance. Adding one is the whole setup: point it at the Kubernetes API and link the observability backends you run.
Add a cluster
Apps → KubeMeridian → Clusters → Add cluster, then fill the one-step form:
- Kubernetes API server URL — e.g.
https://kubernetes.default.svcfor an in-cluster Grafana. - ServiceAccount bearer token — a read-only token (see Deploy for the RBAC). Stored encrypted in
secureJsonData; injected server-side. - Skip TLS verify — on for the in-cluster API certificate.
- Metrics / Logs / Traces — pick your Prometheus, Loki and Tempo datasources.
You can also set these later in the cluster's config (Edit).
Connections
Each cluster links its datasources by UID (stable across renames):
| Connection | Powers | Required? |
|---|---|---|
| Metrics (Prometheus) | all dashboards, cost, RED | for dashboards |
| Logs (Loki) | the Logs page + inline pod logs | optional |
| Traces (Tempo) | the Traces page + service graph | optional |
The Clusters page shows a per-cluster checklist — green chips for what's wired, muted for what's missing — plus live health (nodes ready / pods / namespaces).
Cost & traffic
In the cluster config you also choose:
- Cost (FinOps) — toggle OpenCost present to use real hourly cost metrics, or leave it off to estimate from requests using a configurable price list (
$/vCPU-hour,$/GiB-hour). See Cost & efficiency. - Traffic / RED source — the ingress/mesh stack that exposes your request metrics (Istio, NGINX, Traefik, …) or Auto-detect it. See Traffic stack profiles.
TIP
Legacy installs that used the old single "Prometheus instance" field keep working — it's migrated to the metrics link automatically on read.
Next: Topology & nodes.