varie

add-on/monitoring/files/05-alertmanager-config.yaml

@@ -0,0 +1,62 @@
+# =====================================================================
+# AlertmanagerConfig — routing alert per severità
+# Applica con: kubectl apply -f 05-alertmanager-config.yaml
+#
+# Richiede i Secret con le credenziali (webhook Slack, integration key
+# PagerDuty) creati separatamente, es.:
+#
+#   kubectl create secret generic alertmanager-slack \
+#     -n monitoring --from-literal=url='https://hooks.slack.com/services/XXX/YYY/ZZZ'
+#
+#   kubectl create secret generic alertmanager-pagerduty \
+#     -n monitoring --from-literal=serviceKey='YOUR_PAGERDUTY_INTEGRATION_KEY'
+# =====================================================================
+
+apiVersion: monitoring.coreos.com/v1alpha1
+kind: AlertmanagerConfig
+metadata:
+  name: routing-config
+  namespace: monitoring
+  labels:
+    # deve matchare alertmanagerConfigSelector del chart
+    # (di default kube-prometheus-stack seleziona per release label)
+    release: kube-prometheus-stack
+spec:
+  route:
+    groupBy: ["alertname", "namespace"]
+    groupWait: 30s
+    groupInterval: 5m
+    repeatInterval: 4h
+    receiver: "slack-default"
+    routes:
+      - matchers:
+          - name: severity
+            value: critical
+            matchType: "="
+        receiver: "pagerduty-critical"
+        continue: true   # invia ANCHE a slack-default per visibilità
+      - matchers:
+          - name: severity
+            value: warning
+            matchType: "="
+        receiver: "slack-default"
+
+  receivers:
+    - name: "slack-default"
+      slackConfigs:
+        - apiURL:
+            name: alertmanager-slack
+            key: url
+          channel: "#alerts-k8s"
+          sendResolved: true
+          title: '{{ "{{" }} .CommonAnnotations.summary {{ "}}" }}'
+          text: '{{ "{{" }} .CommonAnnotations.description {{ "}}" }}'
+
+    - name: "pagerduty-critical"
+      pagerdutyConfigs:
+        - serviceKey:
+            name: alertmanager-pagerduty
+            key: serviceKey
+          severity: "critical"
+          description: '{{ "{{" }} .CommonAnnotations.summary {{ "}}" }}'
+          sendResolved: true