Skip to content

Observability

Charmarr fleet dashboard

Charmarr is designed from the ground up to integrate with the Canonical Observability Stack (COS). Every charm bundles its own Prometheus exporter, a curated Grafana dashboard, and alert rules. Wire it to COS and you get metrics, logs, traces, and per-app dashboards out of the box. No exporter hunting. No dashboard scavenging. No glue code.

COS is itself a charmed, open-source observability bundle. It ships:

  • Mimir for metrics
  • Loki for logs
  • Tempo for traces
  • Grafana for dashboards and datasources
  • AlertManager for routing alerts
  • OpenTelemetry Collector as the telemetry ingress

Charmarr plugs into all five via standard Juju relations.

  • Setup


    One Terraform variable, or a handful of juju integrate commands.

    Read more

  • Dashboards


    Per-charm Grafana dashboards shipped with every charm. Zero ops.

    Read more

  • Crowsnest


    Workloadless fleet observability charm: relation graph, SLOs, alerts.

    Read more

What you get out of the box

Concern Vanilla K8s Charmarr + COS
Per-app exporter Find, deploy, configure manually Bundled in the charm sidecar
Per-app dashboard Hunt on Grafana.com, import, tweak Ships with the charm, auto-imported
Alert rules Author or copy-paste Curated per charm, auto-published
Metrics pipeline Wire Prometheus or OTel by hand One juju integrate
Logs pipeline Promtail, Fluent, or OTel by hand One juju integrate
Traces pipeline OTel collector by hand One juju integrate
Fleet topology graph Build it yourself Crowsnest does it

Hours of ops work collapsed into a relation.

Only COS is officially supported

These docs cover COS only. Other backends are not documented and not officially supported. If you want to forward telemetry to a non-COS backend (Grafana Cloud, Signoz, Datadog, VictoriaMetrics, any OTLP target), it is technically possible by deploying opentelemetry-collector-k8s in the charmarr model and configuring it as a forwarder. You give up auto-imported dashboards and auto-loaded alert rules. Experiments are welcome and PRs adding documented support for additional backends are encouraged.

COS itself is not covered here

Deploying and operating COS is out of scope. See the official COS documentation for installation, scaling, retention tuning, and high-availability patterns.