Dashboards

KubeMeridian bundles curated Prometheus dashboards. They're computed inline from raw kube-state-metrics, node-exporter and cAdvisor — they do not depend on the kubernetes-mixin recording rules, so they work on a vanilla kube-prometheus-stack. Every panel binds to the cluster's linked metrics datasource via a $PromDs variable.

Dashboard	What it shows
Cluster Overview	Capacity vs allocatable vs requests vs usage (CPU/mem), node readiness, pod phases, top namespaces, network/disk
Namespace Overview	Per-namespace usage, quotas, pods-by-phase, top workloads, restarts
Workload Health	CrashLoopBackOff, OOMKilled, ImagePull errors, pending / not-ready pods, restart rates, near-limit & throttling hotspots
Storage	PVC/inodes usage, used-vs-capacity, fill forecast (predict_linear), PV phases
Cost & Efficiency	Estimated/real cost, waste, right-sizing — see Cost
Control Plane	apiserver rate/errors/latency, etcd, scheduler, controller-manager (empty on managed clusters that hide these)
Networking	Pod/node network + Istio mesh traffic + ingress-controller metrics
Node / Pod / Deployment / StatefulSet / DaemonSet	The classic per-object resource dashboards

Graceful degradation

Panels whose metrics aren't present in your environment render as No data rather than erroring — e.g. control-plane panels on EKS/GKE/AKS/DOKS, or the ingress-nginx section when you run a different controller. Each such dashboard carries a short note explaining the dependency.

Robustness

• $__rate_interval everywhere, so rates stay correct across scrape intervals and zoom.
• A $cluster selector (=~, default All) that also matches series without a cluster label — so single-cluster Prometheus setups just work.