0015. Use kube-prometheus-stack for metrics and dashboards¶
- Status: Accepted
- Date: 2026-09-26
Context¶
Throughput, consumer lag, end-to-end latency and resource use must be measured to prove the billion-event goal and to operate the platform. Strimzi exposes Prometheus metrics. The log pipeline is a separate, open decision (Phase 2).
Options¶
| Option | License | Pros | Cons |
|---|---|---|---|
| kube-prometheus-stack (Prometheus Operator, Alertmanager, Grafana) | Apache-2.0 (Grafana AGPL-3.0) | Standard, ServiceMonitor CRDs, Strimzi examples | Grafana is copyleft (fine for self-hosted use) |
| VictoriaMetrics stack | Apache-2.0 | Lighter, efficient | Fewer ready examples for Strimzi |
| Managed SaaS | — | No ops | Not local, not open source |
Decision¶
Use kube-prometheus-stack with Strimzi's metrics and dashboards; SLOs and alert routing are defined in later phases.
Consequences¶
- Grafana is AGPL-3.0: used unmodified as a service, which is acceptable; noted for awareness.
- Prometheus retention is sized per profile to respect the disk budget.