-
Notifications
You must be signed in to change notification settings - Fork 0
Description
Goal
Get metrics working without the mess of a bundled stack chart.
Approach
App A: Operator/CRDs
App B: Prometheus instance + scrape configuration (via ServiceMonitors/PodMonitors)
Exporters as separate apps if needed (kube-state-metrics, node exporter), or keep only what you actually use.
Scope / Tasks
Create Argo CD Application for Prometheus Operator:
argocd/apps/platform/prometheus-operator.yml
Chart: Prometheus Operator chart (choose the canonical community chart or Prometheus-community operator chart; do not install kube-prometheus-stack)
Ensure CRDs handled appropriately for Argo sync (explicit CRD install setting if chart supports it)
Create Argo CD Application for Prometheus instance:
argocd/apps/platform/prometheus.yml
Use either:
the same chart if it supports deploying a Prometheus CR cleanly, or
a small k8s manifest folder under k8s/prometheus/ containing the Prometheus CR + Service + RBAC
Persistence:
Use a Retain storage class for Prometheus PVC
Exporters/integrations:
Decide minimal set:
kube-state-metrics (recommended)
node-exporter (optional if Alloy covers your needs)
If used, deploy them as separate Argo apps under platform.
Acceptance Criteria
Prometheus is scraping Kubernetes targets (at least kube-state-metrics)
Prometheus has persistent storage using a Retain class
Operator and instance are independently upgradeable
No kube-prometheus-stack chart is used