# Prometheus

Argo CD installs the upstream `kube-prometheus-stack` Helm chart at the
version selected in `bootstrap/prometheus.yaml`. `values.yaml` contains only
this cluster's intentional settings. The chart owns Prometheus, Alertmanager,
their supporting workloads, and chart-provided monitoring resources.

The `resources/` directory contains cluster-specific alert rules that are not
chart defaults. It does not contain application workload templates. Immich's
host metrics endpoints are configured as static scrape targets in `values.yaml`;
this avoids depending on legacy Endpoints objects, which this Argo CD instance
intentionally excludes from reconciliation.

The chart's CRDs are skipped during normal sync because the existing CRDs are
cluster-wide and their large schemas exceed client-side apply annotation limits.
They remain installed; evaluate the chart's documented CRD upgrade path
separately before a future chart upgrade.

Prometheus TSDB uses the existing 50Gi `rook-ceph-block` claim. During adoption,
preserve claim
`prometheus-prometheus-kube-prometheus-prometheus-db-prometheus-prometheus-kube-prometheus-prometheus-0`
and PV `pvc-dfeec228-10e1-42f2-af69-ebcac00f5ddc`; do not delete or recreate it.

`kubeadm` control-plane metrics listener and host firewall settings are tracked
separately under `cluster-config/kubeadm-monitoring/`. Argo CD does not manage
those host files or static-pod manifests.