ankra-observability

name

ankra-observability

description

Deploy monitoring and logging on a Kubernetes cluster through Ankra (Prometheus, Grafana, Loki, Promtail) and wire external Prometheus-compatible metrics sources (Grafana Cloud, Amazon Managed Prometheus, Google Cloud Managed Prometheus, Thanos, VictoriaMetrics). Use when the user wants monitoring, metrics, dashboards, or logging on a Kubernetes cluster in an Ankra environment, or to connect a metrics source to Ankra.

Ankra Observability

Ankra deploys observability as ordinary stacks/addons and can also read from external Prometheus-compatible metrics sources to power dashboards and AI insights.

Monitoring + logging stack

Deploy as a focused stack. Pin versions, create the namespace first:

stacks:
  - name: monitoring
    manifests:
      - name: monitoring-ns
        parents: []
        from_file: "manifests/monitoring-namespace.yaml"
    addons:
      - name: kube-prometheus-stack       # Prometheus + Grafana + Alertmanager
        chart_name: kube-prometheus-stack
        chart_version: 65.1.1
        repository_url: https://prometheus-community.github.io/helm-charts
        namespace: monitoring
        parents:
          - manifest: monitoring-ns
        configuration:
          values: |-
            grafana:
              adminPassword: ""           # set via SOPS-encrypted value, not plaintext
      - name: loki                          # logs
        chart_name: loki
        chart_version: 6.6.4
        repository_url: https://grafana.github.io/helm-charts
        namespace: monitoring
        parents:
          - manifest: monitoring-ns
      - name: promtail                      # log shipping
        chart_name: promtail
        chart_version: 6.16.4
        repository_url: https://grafana.github.io/helm-charts
        namespace: monitoring
        parents:
          - addon: loki

External metrics sources

Instead of (or in addition to) in-cluster Prometheus, connect a Prometheus-compatible source so Ankra reads metrics for dashboards/insights: Grafana Cloud, Amazon Managed Prometheus, Google Cloud Managed Prometheus, Thanos, or VictoriaMetrics. Provide the query endpoint URL and scoped, read-only credentials stored in Ankra.

Rules

• Namespace first, then the stack via parents.
• Pin chart versions.
• No plaintext credentials (Grafana admin password, remote-write tokens) - use SOPS-encrypted values (ankra-sops-secrets).
• Read-only credentials for external metrics sources.
• Right-size retention and resources in values; avoid unbounded storage defaults in production.

Related skills

• ankra-stacks-addons for stack composition.
• ankra-alerts-webhooks to route alerts from this stack to Slack/Teams/PagerDuty with AI analysis.