apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: factorio
  namespace: factorio
  labels:
    app.kubernetes.io/name: factorio
spec:
  selector:
    matchLabels:
      app.kubernetes.io/name: factorio
      app.kubernetes.io/component: metrics
  endpoints:
  - port: metrics
    interval: 30s
    scrapeTimeout: 10s
---
apiVersion: monitoring.coreos.com/v1
kind: PrometheusRule
metadata:
  name: factorio
  namespace: factorio
  labels:
    app.kubernetes.io/name: factorio
spec:
  groups:
  - name: factorio
    rules:
    - alert: FactorioUnavailable
      expr: kube_deployment_status_replicas_available{namespace="factorio",deployment="factorio"}
        < 1 or absent(kube_deployment_status_replicas_available{namespace="factorio",deployment="factorio"})
      for: 5m
      labels:
        severity: warning
      annotations:
        summary: Factorio has no ready server
    - alert: FactorioMetricsUnavailable
      expr: up{namespace="factorio",service="factorio-metrics"} == 0 or absent(up{namespace="factorio",service="factorio-metrics"})
      for: 5m
      labels:
        severity: warning
      annotations:
        summary: Factorio metrics collection is failing
    - alert: FactorioRestarting
      expr: sum by (container) (increase(kube_pod_container_status_restarts_total{namespace="factorio"}[15m]))
        > 2
      for: 5m
      labels:
        severity: warning
      annotations:
        summary: Factorio container is repeatedly restarting
    - alert: FactorioStorageAlmostFull
      expr: kubelet_volume_stats_available_bytes{namespace="factorio",persistentvolumeclaim="factorio-data"}
        / kubelet_volume_stats_capacity_bytes{namespace="factorio",persistentvolumeclaim="factorio-data"}
        < 0.15
      for: 10m
      labels:
        severity: warning
      annotations:
        summary: Factorio save storage has less than 15% free
    - alert: FactorioGameNotResponding
      expr: factorio_server_up{namespace="factorio"} == 0 or absent(factorio_server_up{namespace="factorio"})
      for: 3m
      labels:
        severity: warning
      annotations:
        summary: Factorio is not answering authenticated player queries
    - alert: FactorioPlayerCollectorStale
      expr: time() - factorio_collector_last_poll_timestamp_seconds{namespace="factorio"}
        > 60
      for: 2m
      labels:
        severity: warning
      annotations:
        summary: Factorio player collector has stopped updating