Skip to main content

Observability

Each organization receives its own view of metrics, logs, and dashboards.

Kube-DC includes a multi-tenant observability stack based on Grafana, Mimir, and Loki. When an organization is created, platform controllers provision a Grafana organization, dashboards, a metrics tenant, and log routing. Teams can open Grafana and inspect their workloads without installing a separate monitoring stack.

What every organization gets​

  • A Grafana organization of its own — members sign in with the same SSO they use everywhere else and land in their organization's view.
  • Pre-provisioned dashboards for the organization's projects and workloads — created and maintained by the platform's controllers, not hand-built per team.
  • Metrics collected from the organization's workloads and served from the multi-tenant metrics store, scoped per tenant.
  • Logs from the organization's workloads, routed per tenant and queryable from the same Grafana.
  • Managed-cluster control-plane telemetry: the logs and events of an organization's managed Kubernetes clusters surface in its Grafana — API servers and controllers are watchable without filing a ticket.
Per-Organization observability provisioning and data flowWhen an Organization is created, Kube-DC controllers provision its Grafana organization and dashboards, a Mimir metrics tenant, and Loki log routing. Project workloads and managed-cluster control planes send telemetry to the shared multi-tenant stack. Organization members authenticate with SSO and query only their tenant scope through Grafana.PROVISION · COLLECT · ISOLATE · PRESENTOrganizationtenant lifecycleKube-DC controllersautomatic reconciliationTenant viewGrafana orgdashboardsProject workloadsapps · VMsdatabasesControl planesAPI serverscontrollersSHARED STACK · TENANT-SCOPED DATATelemetry storeMimir metricsLoki logsTenant membersorganization SSODeployment policyRetention and coverage remain operator-defined.
Organization reconciliation provisions Grafana, metrics tenancy and log routing; workloads and managed control planes feed the shared data layer, while tenant-scoped queries return only that organization's data.

Tenant separation​

The stack is shared; the data is not. Metrics and logs are separated per tenant at the data layer — queries from one organization's Grafana reach only that organization's data. The platform operates one observability system for all tenants instead of one stack per team, which is what makes day-one monitoring economically automatic.

The operator's view​

The platform team monitors the platform itself from the same tooling:

  • Cluster health, capacity, networking and component status, with alerting for platform operators.
  • Storage health: Ceph dashboards and alerts for the storage layer, surfaced in the operator's monitoring and the administration console.
  • Coverage and retention are operator-defined — the bundled defaults are a starting point, tuned to your capacity in the architecture review.

Responsibilities​

ConcernYour platform teamTenant
Observability stack operation and capacity✅—
Per-organization provisioning (Grafana org, dashboards, tenants, routing)✅ automatic, by controllers—
Retention and coverage configuration✅—
Watching workload dashboards, acting on application signals—✅
Custom dashboards within the organization's Grafana—✅
Platform alerting and response✅reports suspected platform issues