Alerts

alert: AdressaytaDown
expr: up{job="adressayta-django"}
  == 0
for: 2m
labels:
  severity: critical
annotations:
  summary: Django metrics недоступны

alert: HostDiskSpaceLow
expr: (node_filesystem_avail_bytes{fstype!~"tmpfs|overlay"}
  / node_filesystem_size_bytes) < 0.1
for: 15m
labels:
  severity: warning
annotations:
  summary: 'Диск < 10% на {{ $labels.host }}: {{ $labels.mountpoint }}'

alert: HostHighCPU
expr: 100
  - (avg by (host, instance) (rate(node_cpu_seconds_total{mode="idle"}[5m]))
  * 100) > 90
for: 10m
labels:
  severity: warning
annotations:
  summary: CPU > 90% на {{ $labels.host }}

alert: HostHighMemory
expr: (1
  - node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes) > 0.9
for: 10m
labels:
  severity: warning
annotations:
  summary: RAM > 90% на {{ $labels.host }}

alert: PostgresDown
expr: up{job="postgres"}
  == 0
for: 2m
labels:
  severity: critical
annotations:
  summary: postgres_exporter недоступен

alert: PostgresTooManyConnections
expr: sum
  by (host) (pg_stat_activity_count) > 80
for: 5m
labels:
  severity: warning
annotations:
  summary: Много подключений PostgreSQL на {{ $labels.host }}

alert: RedisDown
expr: up{job="redis"}
  == 0
for: 2m
labels:
  severity: critical
annotations:
  summary: redis_exporter недоступен

alert: TargetDown
expr: up == 0
for: 3m
labels:
  severity: warning
annotations:
  description: '{{ $labels.host }} ({{ $labels.role }}) недоступен для Prometheus'
  summary: 'Target down: {{ $labels.job }} {{ $labels.instance }}'

alert: HomeVpnBackupDown
expr: homevpn_path_up{host="lenovotc",path="server3"}
  == 0
for: 5m
labels:
  severity: warning
annotations:
  summary: Home VPN backup (S3/wgnet2) down

alert: HomeVpnDown
expr: homevpn_ok{host="lenovotc"}
  == 0
for: 2m
labels:
  severity: critical
annotations:
  description: Оба пути wgnet/wgnet2 не пингуются или юниты down
  summary: Home VPN на TC недоступен (ни S1, ни S3)

alert: HomeVpnOnBackup
expr: homevpn_upstream{host="lenovotc",name="server3-BACKUP"}
  == 1
for: 1m
labels:
  severity: warning
annotations:
  summary: Home VPN работает на backup Server3

alert: HomeVpnPrimaryDown
expr: homevpn_path_up{host="lenovotc",path="server1"}
  == 0
for: 3m
labels:
  severity: warning
annotations:
  description: failover должен уйти на server3-BACKUP
  summary: Home VPN primary (S1/wgnet) down

alert: HomeVpnStaleHandshake
expr: homevpn_peer_handshake_age_seconds{host="lenovotc"}
  > 180
for: 3m
labels:
  severity: warning
annotations:
  summary: WG handshake stale на {{ $labels.iface }} ({{ $value }}s)