You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
feat(observability): retire kube-prometheus-stack like the reference platform #387
The reference platform runs no kube-prometheus-stack. Its alerting moved to Coroot inspections, Flux notification-controller Alerts and off-cluster heartbeats (platform docs/dr/alerting.md, "Changed from the old stack"). The template's Coroot profile still keeps kube-prometheus-stack, and three things depend on it:
the platform-critical PrometheusRule (k8s/bases/infrastructure/alerts/, kept as FluxHelmReleaseNotReady by retain-helm-release-alert.yaml);
the Alertmanager Watchdog dead-man path (ALERTMANAGER_HEARTBEAT_URL in bootstrap.yaml);
the ServiceMonitor, PodMonitor and PrometheusRule CRDs that several chart values enable.
Proposed change, after the default-path sub-issue lands
Move HelmRelease failure coverage to the Flux-native path. Add HelmRelease to the production platform-reconciliation Alert's eventSources, then drop the PrometheusRule and k8s/bases/infrastructure/alerts/, removing it from .github/manifest-naming.yamlcr-directories.
Retire the Watchdog heartbeat in favour of the Coroot cluster-heartbeat, including the bootstrap secret handling, validate-bootstrap-workflow.sh and the docs.
Turn off every chart's serviceMonitor, podMonitor and prometheusRule values that depend on the Prometheus operator CRDs. Drop the allowances for the monitoring namespace from the CiliumNetworkPolicies and security exceptions.
Delete k8s/bases/infrastructure/controllers/kube-prometheus-stack, the hetzner store-tsdb-on-hcloud patch and the remove-kps-ui component.
Activity
chore: agent-claim #387 nonce=137520f2efef6a7b61963583a367aa2e
fix(observability): bound the heartbeat preflight extraction and tidy…
test(observability): tighten two checks the review flagged
feat(observability)!: retire kube-prometheus-stack (#425)