Skip to main content
Version: main 🚧

Operate the fleet

This section covers seeing across many clusters at once. It's about collecting metrics and logs from every cluster the platform manages into one place, for the platform admins and operators who run that monitoring. Monitoring the platform's own components belongs to Monitoring in Administer.

Most of the section is Fleet Observability, the Platform-managed feature. Collectors on each cluster send metrics and logs through a platform gateway. The gateway authenticates every write and stamps platform-owned labels that carry the cluster's trusted identity. Its scoped query API limits who can read the results. The rest of the section is older or narrower. It holds monitoring reference architectures that predate Fleet Observability, and a log mapper for clusters on shared nodes.

Collect fleet metrics​

Use Fleet Observability for new fleet-wide metrics deployments. The exceptions depend on how the platform and its clusters run.

IfUse
You're setting up fleet metrics for the first timeFleet Observability. The quickstart deploys the default pipeline end to end with Argo CD.
The platform runs on embedded Kubernetes, which backs the external database and multi-region featuresA monitoring reference architecture. Fleet Observability is turned off automatically in these deployments.
Some clusters are vCluster StandaloneFleet Observability, with collectors you configure manually. The bundled Argo CD template doesn't cover them.
You already run one of the older architecturesCompare it with Fleet Observability in Fleet metrics architectures. The architectures don't share labels, so dashboards and alerts built for one don't carry over.

The reference architectures are operator-maintained, so you maintain them yourself.

Collect fleet logs​

For logs, where a cluster's pods run decides the approach.

If the cluster runs onUse
Private nodesFleet Observability pod log collection, which ships logs through the gateway labeled with the cluster's verified identity.
Shared nodesCentral HostPath Mapper with your own collector, such as Fluentd or Grafana Loki. Pods from several clusters share each node, and the mapper matches log files to the cluster pods that wrote them.

Fleet Observability log collection finds no pod logs on shared nodes without the mapper. Nothing fails, so the gap is easy to miss. For vCluster Standalone, the Docker driver, and the full comparison, see Choose a log collection approach.

Covered in other sections​

  • To monitor a single connected cluster with Prometheus and Grafana, see Single-cluster monitoring.
  • To find platform audit logs and process logs, see Logging.
  • To read the logs of a single cluster, see Logging in the vCluster docs.