v1.29_Core Checker

version: "3"

services:
  prometheus:
    image: prom/prometheus:latest
    container_name: prometheus
    volumes:
      - ./conf:/etc/prometheus
      - prometheus_data:/prometheus
    command:
      - '--config.file=/etc/prometheus/prometheus.yml'
      - '--storage.tsdb.path=/prometheus'
      - '--storage.tsdb.retention.time=5d'
    ports:
      - "9090:9090"

  grafana:
    image: hub.eltex-co.ru/softwlc/eltex-core-checker:latest
    container_name: grafana
    depends_on:
      - prometheus
    ports:
      - 3000:3000
    links:
      - alertmanager:alertmanager
    volumes:
      - grafana_data:/var/lib/grafana

  alertmanager:
    image: prom/alertmanager:latest
    container_name: alertmanager
    volumes:
      - ./conf:/etc/alertmanager/
    ports:
      - 9093:9093
    command:
      - '--config.file=/etc/alertmanager/alert_config.yml'
      - '--storage.path=/alertmanager'
    depends_on:
      - prometheus

  node-exporter:
    hostname: monitoring-server
    image: prom/node-exporter:v1.3.1
    volumes:
        - /proc:/host/proc:ro
        - /sys:/host/sys:ro
        - /:/rootfs:ro
    command:
        - --path.procfs=/host/proc
        - --path.sysfs=/host/sys
        - --collector.filesystem.ignored-mount-points
        - ^/(sys|proc|dev|host|etc|rootfs/var/lib/docker/containers|rootfs/var/lib/docker/overlay2|rootfs/run/docker/netns|rootfs/var/lib/docker/aufs)($$|/)
    ports:
        - 9101:9100

volumes:
  grafana_data:
  prometheus_data:

 scrape_configs:

  - job_name: 'core-monitoring'
    scrape_interval: 5s
    metrics_path: '/metrics'
    static_configs:
      - targets: ['172.17.0.1:8000']

  - job_name: 'monitoring-server'
    scrape_interval: 5s
    metrics_path: 'metrics'
    static_configs:
      - targets: ['172.17.0.1:9100']

  - job_name: 'alertmanager'
    scrape_interval: 5s
    metrics_path: '/metrics'
    static_configs:
     - targets: ['172.17.0.1:9093']

 # Alertmanager configuration
 alerting:
   alertmanagers:
    - scheme: http
      static_configs:
      - targets:
         - alertmanager:9093

 # Load rules once and periodically evaluate them according to the global 'evaluation_interval'.
 rule_files:
   - 'alert.rules'

global:
  smtp_smarthost: 'smtp.yandex.ru:25'
  smtp_from: 'user@yandex.ru'
  smtp_auth_username: 'user@yandex.ru'
  smtp_auth_password: 'password'
  resolve_timeout: 10s

route:
   receiver: 'abon'
   group_interval: 10s

receivers:
  - name: 'abon'
    email_configs:
    - send_resolved: True
      to: 'user_all@yandex.ru'
#    webhook_configs:
#    - send_resolved: True
#      http_config:
#        proxy_url: http://172.17.0.1:8118
#      url: 'http://172.17.0.1/alert/-000000000'

groups:
- name: abon
  rules:
        
  - alert: ems_state_pcrf
    expr: ems_state_pcrf{ems_state_pcrf="false"} > 0
    for: 10s
    annotations:
      summary: "ems_state_pcrf down"
       
  - alert: nbi_state_pcrf
    expr: nbi_state_pcrf{nbi_state_pcrf="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state_pcrf down"
       
  - alert: apb_state_mercury
    expr: apb_state_mercury{apb_state_mercury="false"} > 0
    for: 10s
    annotations:
      summary: "apb_state_mercury down"
       
  - alert: ngw_state_mongo
    expr: ngw_state_mongo{ngw_state_mongo="false"} > 0
    for: 10s
    annotations:
      summary: "ngw_state_mongo down"
       
  - alert: nbi_state
    expr: nbi_state{nbi_state="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state down"
       
  - alert: nbi_state_ems
    expr: nbi_state_ems{nbi_state_ems="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state_ems down"
       
  - alert: doors_state_mysql
    expr: doors_state_mysql{doors_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "doors_state_mysql down"
       
  - alert: nbi_state_mysql
    expr: nbi_state_mysql{nbi_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state_mysql down"
       
  - alert: portal_state_mercury
    expr: portal_state_mercury{portal_state_mercury="false"} > 0
    for: 10s
    annotations:
      summary: "portal_state_mercury down"
       
  - alert: nbi_state_auth_service
    expr: nbi_state_auth_service{nbi_state_auth_service="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state_auth_service down"
       
  - alert: mac_checker_state
    expr: mac_checker_state{mac_checker_state="false"} > 0
    for: 10s
    annotations:
      summary: "mac_checker_state down"
       
  - alert: portal_state
    expr: portal_state{portal_state="false"} > 0
    for: 10s
    annotations:
      summary: "portal_state down"
       
  - alert: pcrf_state
    expr: pcrf_state{pcrf_state="false"} > 0
    for: 10s
    annotations:
      summary: "pcrf_state down"
       
  - alert: prcf_state_nbi
    expr: prcf_state_nbi{prcf_state_nbi="false"} > 0
    for: 10s
    annotations:
      summary: "prcf_state_nbi down"
       
  - alert: nbi_state_mongo
    expr: nbi_state_mongo{nbi_state_mongo="false"} > 0
    for: 10s
    annotations:
      summary: "nbi_state_mongo down"
       
  - alert: lk_state_nbi
    expr: lk_state_nbi{lk_state_nbi="false"} > 0
    for: 10s
    annotations:
      summary: "lk_state_nbi down"
       
  - alert: lk_state_auth_service
    expr: lk_state_auth_service{lk_state_auth_service="false"} > 0
    for: 10s
    annotations:
      summary: "lk_state_auth_service down"
       
  - alert: pcrf_state_mysql
    expr: pcrf_state_mysql{pcrf_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "pcrf_state_mysql down"
       
  - alert: mercury_state
    expr: mercury_state{mercury_state="false"} > 0
    for: 10s
    annotations:
      summary: "mercury_state down"
       
  - alert: portal_state_mysql
    expr: portal_state_mysql{portal_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "portal_state_mysql down"
       
  - alert: mercury_state_mysql
    expr: mercury_state_mysql{mercury_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "mercury_state_mysql down"
 
  - alert: apb_state
    expr: apb_state{apb_state="false"} > 0
    for: 10s
    annotations:
      summary: "apb down"
       
  - alert: doors_state
    expr: doors_state{doors_state="false"} > 0
    for: 10s
    annotations:
      summary: "doors_state down"
       
  - alert: pcrf_state_mongo
    expr: pcrf_state_mongo{pcrf_state_mongo="false"} > 0
    for: 10s
    annotations:
      summary: "pcrf_state_mongo down"
       
  - alert: radius_state
    expr: radius_state{radius_state="false"} > 0
    for: 10s
    annotations:
      summary: "radius_state down"
       
  - alert: auth_service_state
    expr: auth_service_state{auth_service_state="false"} > 0
    for: 10s
    annotations:
      summary: "auth_service_state down"
       
  - alert: disconnect_state
    expr: disconnect_state{disconnect_state="false"} > 0
    for: 10s
    annotations:
      summary: "disconnect_state down"
       
  - alert: lk_state_mongo
    expr: lk_state_mongo{lk_state_mongo="false"} > 0
    for: 10s
    annotations:
      summary: "lk_state_mongo down"
       
  - alert: bob_state
    expr: bob_state{bob_state="false"} > 0
    for: 10s
    annotations:
      summary: "bob_state down"
       
  - alert: disconnect_state_pcrf
    expr: disconnect_state_pcrf{disconnect_state_pcrf="false"} > 0
    for: 10s
    annotations:
      summary: "disconnect_state_pcrf down"
 
  - alert: bob_state_mysql
    expr: bob_state_mysql{bob_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "bob_state_mysql down"
       
  - alert: cp_state_auth_service
    expr: cp_state_auth_service{cp_state_auth_service="false"} > 0
    for: 10s
    annotations:
      summary: "cp_state_auth_service down"
       
  - alert: cp_state_mysql
    expr: cp_state_mysql{cp_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "cp_state_mysql down"
       
  - alert: mac_checker_state_ems
    expr: mac_checker_state_ems{mac_checker_state_ems="false"} > 0
    for: 10s
    annotations:
      summary: "mac_checker_state_ems down"
       
  - alert: ngw_state_smpp
    expr: ngw_state_smpp{ngw_state_smpp="false"} > 0
    for: 10s
    annotations:
      summary: "ngw_state_smpp down"
       
  - alert: ngw_state_mail
    expr: ngw_state_mail{ngw_state_mail="false"} > 0
    for: 10s
    annotations:
      summary: "ngw_state_mail down"
       
  - alert: ems_state
    expr: ems_state{ems_state="false"} > 0
    for: 10s
    annotations:
      summary: "ems_state down"
       
  - alert: ems_state_mysql
    expr: ems_state_mysql{ems_state_mysql="false"} > 0
    for: 10s
    annotations:
      summary: "ems_state_mysql down"