Browsing metrics

The Metrics page has two modes, toggled top-right:

  • Explorer — interactive ad-hoc plotting. Pick a metric, scope it with labels, plot it.
  • Analytics — a pre-built infrastructure summary for the currently-selected source (KPI tiles + Node Health + Pod/Container Analytics tables).

Both modes share the source picker (left rail) and the time picker. They serve different needs:

NeedUse
Plot one specific metric over time, save as alert.Explorer
Quick health snapshot for a source — top CPU consumer, restart count, pod-by-pod table.Analytics
Build a reusable layout for the whole team.Dashboards

Source scope

Metrics is per-source. The source switcher (left rail) needs a single source selected — Analytics does not aggregate across sources.

Analytics needs an agent-backed source

Analytics is built from host and node metrics the agent collects, so it works for Kubernetes, Docker and Linux server / VM sources. A Linux host emits the same node_* series under the same names as the Kubernetes node collector, which is why one view covers all three.

It does not work for a Service / OpenTelemetry app source. Those push their own application metrics via OTLP and no agent reports the machine, so there is no node health to summarise. Selecting one shows an explicit "not supported for this source type" banner rather than empty cards — you are told why, not left to guess whether data is missing.

Explorer works for every source type, including Service. Anything ingested can be plotted; only the pre-built infrastructure summary needs an agent.

One exception inside Analytics: the certificate cards are Kubernetes-only. Certificate posture is read from Ingress and Secret objects, which have no equivalent on a bare host or a Docker daemon.

See also