monitoring: add Grafana alloy and Prometheus kube-state-metrics

2026-09-12 14:07:26 +00:00
parent c44dec8900
commit 719b906a53
14 changed files with 828 additions and 0 deletions
@@ -0,0 +1,91 @@
+controller:
+  nameOverride: "alloy-node-scrapper"
+  volumes:
+    extra:
+      # requires host journal configuration storage = volatile
+      - name: run-log-journal
+        hostPath:
+          path: /run/log/journal
+      # Log scrapers (e.g., Alloy/Loki) read this ID from the journal metadata and use it to separate log streams per machine.
+      - name: etc-machine-id
+        hostPath:
+          path: /etc/machine-id
+      - name: rootfs
+        hostPath:
+          path: /
+      - name: proc
+        hostPath:
+          path: /proc
+      - name: sys
+        hostPath:
+          path: /sys
+      - name: dev
+        hostPath:
+          path: /dev
+
+  tolerations:
+    - key: "node-role.kubernetes.io/control-plane"
+      operator: "Exists"
+      effect: "NoSchedule"
+    - key: "infra.limbosolutions.com/dedicated"
+      operator: "Exists"
+      effect: "NoSchedule"
+
+alloy:
+  mounts:
+    varlog: true  # Mounts /var/log from the host (persistent journal)
+    extra:
+      # requires host journal configuration storage = volatile
+      - name: run-log-journal
+        mountPath: /run/log/journal
+        readOnly: true
+      # required 
+      - name: etc-machine-id
+        mountPath: /etc/machine-id
+        readOnly: true
+      - name: rootfs
+        mountPath: /host
+        readOnly: true
+      - name: proc
+        mountPath: /host/proc
+        readOnly: true
+      - name: sys
+        mountPath: /host/sys
+        readOnly: true
+      - name: dev
+        mountPath: /host/dev
+        readOnly: true 
+  # https://grafana.com/docs/alloy/latest/collect/logs-in-kubernetes/
+  configMap:
+    create: false
+    name: alloy-node-scraper 
+    key: config.alloy
+  extraEnv:
+    - name: NODE_IP
+      valueFrom:
+        fieldRef:
+          fieldPath: status.hostIP
+    - name: CLUSTER_NAME
+      value: casa
+    - name: JOB
+      value: node-exporter
+    - name: CONTAINER
+      value: node-exporter
+    - name: NS
+      value: monitoring
+    - name: POD
+      value: alloy-node-scraper
+    - name: SERVICE
+      value: alloy-node-scraper
+    - name: ZONE
+      value: casa-vlan
+    - name: LOKI_URL
+      valueFrom:
+        secretKeyRef:
+          name: alloy-global
+          key: loki_url
+    - name: PROMETHEUS_URL
+      valueFrom:
+        secretKeyRef:
+          name: alloy-global
+          key: prometheus_url