Alert formatting and swap threshold

monitoring/alertmanager/alertmanager.yml

@@ -15,9 +15,11 @@ receivers:
         chat_id: -1003779247128
         message: |-
           {{ range .Alerts }}
-          *{{ .Labels.alertname }}* ({{ .Labels.severity }})
+          {{ if eq .Labels.severity "critical" }}🔴{{ else }}⚠️{{ end }} *{{ .Labels.alertname }}* ({{ .Labels.severity }})
           {{ .Annotations.summary }}
           {{ .Annotations.description }}
+          Instance: {{ .Labels.instance }}
+          Details: {{ .GeneratorURL }}
           {{ end }}
         parse_mode: Markdown
         disable_notifications: false

monitoring/blackbox/blackbox.yml

@@ -0,0 +1,19 @@
+modules:
+  http_2xx:
+    prober: http
+    timeout: 5s
+    http:
+      valid_http_versions: ["HTTP/1.1", "HTTP/2.0"]
+      valid_status_codes: []
+      method: GET
+      follow_redirects: true
+      preferred_ip_protocol: "ip4"
+  http_5xx:
+    prober: http
+    timeout: 5s
+    http:
+      valid_http_versions: ["HTTP/1.1", "HTTP/2.0"]
+      valid_status_codes: [500, 501, 502, 503, 504]
+      method: GET
+      follow_redirects: true
+      preferred_ip_protocol: "ip4"

monitoring/docker-compose.yml

@@ -77,6 +77,14 @@ services:
     ports:
       - "9633:9633"
 
+  blackbox-exporter:
+    image: prom/blackbox-exporter:v0.25.0
+    restart: unless-stopped
+    volumes:
+      - ./blackbox/blackbox.yml:/etc/blackbox_exporter/config.yml:ro
+    ports:
+      - "9115:9115"
+
 volumes:
   prometheus-data:
   grafana-data:

monitoring/prometheus/prometheus.yml

@@ -30,3 +30,55 @@ scrape_configs:
   - job_name: revproxy-node
     static_configs:
       - targets: ["10.8.0.1:9100"]
+
+  - job_name: blackbox-http
+    metrics_path: /probe
+    params:
+      module: [http_2xx]
+    static_configs:
+      - targets:
+          - https://s3.g4st3r.ru
+          - https://minio.g4st3r.ru
+          - https://cloud.g4st3r.ru
+          - https://grafana.g4st3r.ru
+          - https://prometheus.g4st3r.ru
+          - https://alert.g4st3r.ru
+          - https://photos.g4st3r.ru
+          - https://home.g4st3r.ru
+          - https://gitlab.g4st3r.ru
+          - https://openclaw.g4st3r.ru
+          - https://git.g4st3r.ru
+          - https://bw.g4st3r.ru
+    relabel_configs:
+      - source_labels: [__address__]
+        target_label: __param_target
+      - source_labels: [__param_target]
+        target_label: instance
+      - target_label: __address__
+        replacement: blackbox-exporter:9115
+
+  - job_name: blackbox-5xx
+    metrics_path: /probe
+    params:
+      module: [http_5xx]
+    static_configs:
+      - targets:
+          - https://s3.g4st3r.ru
+          - https://minio.g4st3r.ru
+          - https://cloud.g4st3r.ru
+          - https://grafana.g4st3r.ru
+          - https://prometheus.g4st3r.ru
+          - https://alert.g4st3r.ru
+          - https://photos.g4st3r.ru
+          - https://home.g4st3r.ru
+          - https://gitlab.g4st3r.ru
+          - https://openclaw.g4st3r.ru
+          - https://git.g4st3r.ru
+          - https://bw.g4st3r.ru
+    relabel_configs:
+      - source_labels: [__address__]
+        target_label: __param_target
+      - source_labels: [__param_target]
+        target_label: instance
+      - target_label: __address__
+        replacement: blackbox-exporter:9115

monitoring/rules/base-alerts.yml

@@ -47,13 +47,13 @@ groups:
           description: "MemAvailable < 8%"
 
       - alert: SwapUsageHigh
-        expr: (1 - (node_memory_SwapFree_bytes / node_memory_SwapTotal_bytes)) > 0.30
-        for: 15m
+        expr: (1 - (node_memory_SwapFree_bytes / node_memory_SwapTotal_bytes)) > 0.50
+        for: 20m
         labels:
           severity: warning
         annotations:
           summary: "Swap usage high on {{ $labels.instance }}"
-          description: "Swap usage > 30%"
+          description: "Swap usage > 50%"
 
       - alert: DiskFull
         expr: (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay"} / node_filesystem_size_bytes{fstype!~"tmpfs|overlay"}) < 0.10
@@ -65,22 +65,22 @@ groups:
           description: "<10% free on {{ $labels.mountpoint }}"
 
       - alert: DiskUsageHigh
-        expr: (1 - (node_filesystem_avail_bytes{mountpoint="/",fstype!~"tmpfs|overlay"} / node_filesystem_size_bytes{mountpoint="/",fstype!~"tmpfs|overlay"})) > 0.85
+        expr: (1 - (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay",mountpoint!~"/boot/efi"} / node_filesystem_size_bytes{fstype!~"tmpfs|overlay",mountpoint!~"/boot/efi"})) > 0.85
         for: 15m
         labels:
           severity: warning
         annotations:
           summary: "Disk usage >85% on {{ $labels.instance }}"
-          description: "Root filesystem usage above 85%"
+          description: "{{ $labels.mountpoint }} usage above 85%"
 
       - alert: DiskUsageCritical
-        expr: (1 - (node_filesystem_avail_bytes{mountpoint="/",fstype!~"tmpfs|overlay"} / node_filesystem_size_bytes{mountpoint="/",fstype!~"tmpfs|overlay"})) > 0.95
+        expr: (1 - (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay",mountpoint!~"/boot/efi"} / node_filesystem_size_bytes{fstype!~"tmpfs|overlay",mountpoint!~"/boot/efi"})) > 0.95
         for: 5m
         labels:
           severity: critical
         annotations:
           summary: "Disk usage >95% on {{ $labels.instance }}"
-          description: "Root filesystem usage above 95%"
+          description: "{{ $labels.mountpoint }} usage above 95%"
 
       - alert: DockerContainerDown
         expr: time() - container_last_seen > 120
@@ -90,3 +90,21 @@ groups:
         annotations:
           summary: "Container down"
           description: "A container has not been seen for >2m"
+
+      - alert: HttpEndpointDown
+        expr: probe_success{job="blackbox-http"} == 0
+        for: 2m
+        labels:
+          severity: critical
+        annotations:
+          summary: "HTTP endpoint down: {{ $labels.instance }}"
+          description: "Blackbox probe failed"
+
+      - alert: HttpEndpoint5xx
+        expr: probe_success{job="blackbox-5xx"} == 1
+        for: 1m
+        labels:
+          severity: critical
+        annotations:
+          summary: "HTTP 5xx detected: {{ $labels.instance }}"
+          description: "Endpoint returned 5xx"