Observability Stack
The standard implementation profile, ok-observability-standard, composes open-source components to satisfy the Observability Capability Contract v1.
Standard implementation
| Area | Reference components | Contract purpose |
|---|---|---|
| Metrics | Prometheus, kube-state-metrics, node-exporter | Collect platform and workload measurements |
| Dashboards | Grafana | Provide cluster-local dashboards and data-source queries |
| Alerting | Prometheus rules, Alertmanager | Route actionable alerts to a configured receiver |
| Logs | OpenSearch and a log collector | Collect and search logs under a declared policy |
The underlying kube-prometheus-stack provides much of the metrics, dashboard and alerting stack. The contract names outcomes, not permanent product dependencies. An envelope-specific profile may substitute components while preserving, or explicitly declaring limitations to, contractual guarantees.
Why local observability?
A local stack can continue operating when connectivity to management or other clusters is unavailable. This is especially relevant to sovereign environments and air-gapped deployments. It also means each cluster must budget for local compute, storage and retention.
OpenSearch is the standard profile's log-search implementation. Its richer search capabilities are an implementation choice, not a required backend in the capability contract.
Source documentation
Not covered by v1: Cross-cluster federation or long-term aggregation, tracing, profiling, and native OTLP ingestion. Those require separate implementation or future contract extensions.