feat: add demo alert rules

.env.example

@@ -1,2 +1,10 @@
 LOKI_PORT=3100
 GRAFANA_PORT=4859
+GRAFANA_ALERT_RECEIVERS=desenvolvimento@neiaac.com,direcao@neiaac.com
+GF_SMTP_ENABLED=true
+GF_SMTP_FROM_ADDRESS=no-reply@neiaac.com
+GF_SMTP_FROM_NAME=no-reply
+GF_SMTP_HOST=email-smtp.eu-north-1.amazonaws.com:587
+GF_SMTP_STARTTLS_POLICY=OpportunisticStartTLS
+GF_SMTP_USER=user
+GF_SMTP_PASSWORD=password

docker-compose.yaml

@@ -73,6 +73,8 @@ services:
     volumes:
       - ./grafana-datasources.yaml:/etc/grafana/provisioning/datasources/datasources.yaml
       - ./grafana-dashboards.yaml:/etc/grafana/provisioning/dashboards/dashboards.yaml
+      - ./grafana-alert-triggers.yaml:/etc/grafana/provisioning/alerting/triggers.yaml
+      - ./grafana-alert-contacts.yaml:/etc/grafana/provisioning/alerting/contacts.yaml
       - ./grafana-metrics-dashboard.json:/var/lib/grafana/dashboards/metrics.json
       - ./grafana-logs-dashboard.json:/var/lib/grafana/dashboards/logs.json
       - grafana-data:/var/lib/grafana

grafana-alert-contacts.yaml

@@ -0,0 +1,10 @@
+apiVersion: 1
+
+contactPoints:
+  name: grafana-default-email
+  receivers:
+    - uid: grafana-default-email
+      type: email
+      settings:
+        addresses: ${GRAFANA_ALERT_RECEIVER}
+        singleEmail: true

grafana-alert-triggers.yaml

@@ -0,0 +1,102 @@
+apiVersion: 1
+
+groups:
+  name: Overload
+  folder: metrics
+  interval: 60s
+  rules:
+    - uid: cpu
+      title: CPU Usage
+      condition: A
+      data:
+        - refId: A
+          datasourceUid: "prometheus"
+          model:
+            conditions:
+              - evaluator:
+                  params:
+                    - 50
+                  type: gt
+                operator:
+                  type: and
+                query:
+                  params:
+                    - "100 * (1 - avg(rate(node_cpu_seconds_total{mode='idle', job='node'}[5m])) by (instance))"
+                reducer:
+                  type: last
+                type: query
+            datasource:
+              type: prometheus
+              uid: "prometheus"
+            expression: 1==0
+            intervalMs: 1000
+            maxDataPoints: 43200
+            refId: A
+            type: math
+      noDataState: Alerting
+      execErrState: Alerting
+      for: 60s
+
+    - uid: disk_usage
+      title: Disk Usage
+      condition: A
+      data:
+        - refId: A
+          datasourceUid: "prometheus"
+          model:
+            conditions:
+              - evaluator:
+                  params:
+                    - 90
+                  type: gt
+                operator:
+                  type: and
+                query:
+                  params:
+                    - "100 * (1 - avg(rate(node_filesystem_free_bytes{fstype!='tmpfs', job='node'}[5m])) by (instance) / avg(rate(node_filesystem_size_bytes{fstype!='tmpfs', job='node'}[5m])) by (instance))"
+                reducer:
+                  type: last
+                type: query
+            datasource:
+              type: prometheus
+              uid: "prometheus"
+            expression: 1==0
+            intervalMs: 1000
+            maxDataPoints: 43200
+            refId: A
+            type: math
+      noDataState: Alerting
+      execErrState: Alerting
+      for: 60s
+
+    - uid: ram_usage
+      title: RAM Usage
+      condition: A
+      data:
+        - refId: A
+          datasourceUid: "prometheus"
+          model:
+            conditions:
+              - evaluator:
+                  params:
+                    - 50
+                  type: gt
+                operator:
+                  type: and
+                query:
+                  params:
+                    - "100 * (1 - avg(rate(node_memory_MemFree_bytes{job='node'}[5m])) by (instance) / avg(rate(node_memory_MemTotal_bytes{job='node'}[5m])) by (instance))"
+                reducer:
+                  type: last
+                type: query
+            datasource:
+              type: prometheus
+              uid: "prometheus"
+            expression: 1==0
+            intervalMs: 1000
+            maxDataPoints: 43200
+            refId: A
+            type: math
+      noDataState: Alerting
+      execErrState: Alerting
+      for: 60s

grafana-dashboards.yaml

@@ -4,7 +4,6 @@ apiVersion: 1
 
 providers:
   - name: Dashboards
-    orgId: 1
     type: file
 
     updateIntervalSeconds: 10