add smartctl exporter

kubernetes/apps/observability/kustomization.yaml

@@ -11,6 +11,7 @@ resources:
   - ./grafana/ks.yaml
   - ./node-exporter/ks.yaml
   - ./prometheus-operator-crds/ks.yaml
+  - ./smartctl-exporter/ks.yaml
   - ./unpoller/ks.yaml
   - ./vector-agent/ks.yaml
   - ./vector-aggregator/ks.yaml

kubernetes/apps/observability/smartcl-exporter/app/helmrelease.yaml

@@ -0,0 +1,30 @@
+---
+# yaml-language-server: $schema=https://ks.hsn.dev/helm.toolkit.fluxcd.io/helmrelease_v2.json
+apiVersion: helm.toolkit.fluxcd.io/v2
+kind: HelmRelease
+metadata:
+  name: &app smartctl-exporter
+spec:
+  interval: 30m
+  chart:
+    spec:
+      chart: prometheus-smartctl-exporter
+      version: 0.12.0
+      sourceRef:
+        kind: HelmRepository
+        name: prometheus-community
+        namespace: flux-system
+  install:
+    remediation:
+      retries: 3
+  upgrade:
+    cleanupOnFail: true
+    remediation:
+      strategy: rollback
+      retries: 3
+  values:
+    fullnameOverride: *app
+    serviceMonitor:
+      enabled: true
+    prometheusRules:
+      enabled: false

kubernetes/apps/observability/smartcl-exporter/app/kustomization.yaml

@@ -0,0 +1,19 @@
+---
+# yaml-language-server: $schema=https://json.schemastore.org/kustomization
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+resources:
+  - ./helmrelease.yaml
+  - ./prometheusrule.yaml
+configMapGenerator:
+  - name: smartctl-exporter-dashboard
+    files:
+      - ./resources/blesswinsamuel_smartctl.json
+      - ./resources/smartctl_exporter.json
+    options:
+      annotations:
+        kustomize.toolkit.fluxcd.io/substitute: disabled
+      labels:
+        grafana_dashboard: "1"
+generatorOptions:
+  disableNameSuffixHash: true

kubernetes/apps/observability/smartcl-exporter/app/prometheusrule.yaml

@@ -0,0 +1,64 @@
+---
+# yaml-language-server: $schema=https://ks.hsn.dev/monitoring.coreos.com/prometheusrule_v1.json
+apiVersion: monitoring.coreos.com/v1
+kind: PrometheusRule
+metadata:
+  name: smartctl-exporter-rules
+spec:
+  groups:
+    - name: smartctl-exporter.rules
+      rules:
+        - alert: SmartDeviceHighTemperature
+          annotations:
+            summary: Mounted drive {{ $labels.device }} on device {{ $labels.instance }}
+              has a temperature higher than 65°C.
+          expr: smartctl_device_temperature > 65
+          for: 15m
+          labels:
+            severity: critical
+        - alert: SmartDeviceTestFailed
+          annotations:
+            summary: Mounted drive {{ $labels.device }} on device {{ $labels.instance }}
+              did not pass its SMART test.
+          expr: |
+            (
+              smartctl_device_smart_status != 1
+            or
+              smartctl_device_status != 1
+            )
+          for: 15m
+          labels:
+            severity: critical
+        - alert: SmartDeviceCriticalWarning
+          annotations:
+            summary: Mounted drive {{ $labels.device }} on device {{ $labels.instance }}
+              is in a critical state.
+          expr: smartctl_device_critical_warning != 0
+          for: 15m
+          labels:
+            severity: critical
+        - alert: SmartDeviceMediaErrors
+          annotations:
+            summary: Mounted drive {{ $labels.device }} on device {{ $labels.instance }}
+              has media errors.
+          expr: smartctl_device_media_errors{device!~"^nvme.+"} != 0
+          for: 15m
+          labels:
+            severity: critical
+        - alert: SmartDeviceAvailableSpareUnderThreadhold
+          annotations:
+            summary: Device {{ $labels.device }} on instance {{ $labels.instance }}
+              is under available spare threashold.
+          expr: smartctl_device_available_spare_threshold > smartctl_device_available_spare
+          for: 15m
+          labels:
+            severity: critical
+        - alert: SmartDeviceInterfaceSlow
+          annotations:
+            summary: Device {{ $labels.device }} on instance {{ $labels.instance }}
+              interface is slower then it should be.
+          expr: |
+            smartctl_device_interface_speed{speed_type="current"} != on(device, instance, namespace, pod) smartctl_device_interface_speed{speed_type="max"}
+          for: 15m
+          labels:
+            severity: critical

kubernetes/apps/observability/smartcl-exporter/ks.yaml

@@ -0,0 +1,20 @@
+---
+# yaml-language-server: $schema=https://ks.hsn.dev/kustomize.toolkit.fluxcd.io/kustomization_v1.json
+apiVersion: kustomize.toolkit.fluxcd.io/v1
+kind: Kustomization
+metadata:
+  name: &app smartctl-exporter
+  namespace: flux-system
+spec:
+  targetNamespace: observability
+  commonMetadata:
+    labels:
+      app.kubernetes.io/name: *app
+  path: ./kubernetes/apps/observability/smartctl-exporter/app
+  prune: true
+  sourceRef:
+    kind: GitRepository
+    name: theshire
+  wait: false
+  interval: 30m
+  retryInterval: 1m