fluxer/fluxer_devops/signoz/compose.yaml

x-common: &common
  networks:
    - fluxer-shared
  logging:
    options:
      max-size: 50m
      max-file: '3'

x-deploy-base: &deploy_base
  restart_policy:
    condition: on-failure
    delay: 5s
    max_attempts: 3

x-clickhouse-defaults: &clickhouse_defaults
  <<: *common
  image: clickhouse/clickhouse-server:25.5.6
  tty: true
  environment:
    - CLICKHOUSE_SKIP_USER_SETUP=1
  deploy:
    <<: *deploy_base
    labels:
      signoz.io/scrape: 'true'
      signoz.io/port: '9363'
      signoz.io/path: '/metrics'
  healthcheck:
    test:
      - CMD
      - wget
      - --spider
      - -q
      - 0.0.0.0:8123/ping
    interval: 30s
    timeout: 5s
    retries: 3
  ulimits:
    nproc: 65535
    nofile:
      soft: 262144
      hard: 262144

x-zookeeper-defaults: &zookeeper_defaults
  <<: *common
  image: signoz/zookeeper:3.7.1
  user: root
  deploy:
    <<: *deploy_base
    labels:
      signoz.io/scrape: 'true'
      signoz.io/port: '9141'
      signoz.io/path: '/metrics'
  healthcheck:
    test:
      - CMD-SHELL
      - curl -s -m 2 http://localhost:8080/commands/ruok | grep error | grep null
    interval: 30s
    timeout: 5s
    retries: 3

services:
  init-clickhouse:
    <<: *common
    image: clickhouse/clickhouse-server:25.5.6
    command:
      - bash
      - -c
      - |
        version="v0.0.1"
        node_os=$$(uname -s | tr '[:upper:]' '[:lower:]')
        node_arch=$$(uname -m | sed s/aarch64/arm64/ | sed s/x86_64/amd64/)
        echo "Fetching histogram-binary for $${node_os}/$${node_arch}"
        cd /tmp
        wget -O histogram-quantile.tar.gz "https://github.com/SigNoz/signoz/releases/download/histogram-quantile%2F$${version}/histogram-quantile_$${node_os}_$${node_arch}.tar.gz"
        tar -xvzf histogram-quantile.tar.gz
        mv histogram-quantile /var/lib/clickhouse/user_scripts/histogramQuantile
    deploy:
      restart_policy:
        condition: on-failure
    volumes:
      - ./conf/clickhouse/user_scripts:/var/lib/clickhouse/user_scripts:rw

  zookeeper-1:
    <<: *zookeeper_defaults
    environment:
      - ZOO_SERVER_ID=1
      - ALLOW_ANONYMOUS_LOGIN=yes
      - ZOO_AUTOPURGE_INTERVAL=1
      - ZOO_ENABLE_PROMETHEUS_METRICS=yes
      - ZOO_PROMETHEUS_METRICS_PORT_NUMBER=9141
    volumes:
      - zookeeper-1:/bitnami/zookeeper

  clickhouse:
    <<: *clickhouse_defaults
    hostname: clickhouse
    configs:
      - source: clickhouse-config
        target: /etc/clickhouse-server/config.xml
      - source: clickhouse-users
        target: /etc/clickhouse-server/users.xml
      - source: clickhouse-custom-function
        target: /etc/clickhouse-server/custom-function.xml
      - source: clickhouse-cluster
        target: /etc/clickhouse-server/config.d/cluster.xml
    volumes:
      - clickhouse:/var/lib/clickhouse/

  schema-migrator:
    <<: *common
    image: signoz/signoz-schema-migrator:${OTELCOL_TAG:-v0.129.12}
    entrypoint: sh
    command:
      - -c
      - /signoz-schema-migrator sync --dsn=tcp://clickhouse:9000 --up= && /signoz-schema-migrator async --dsn=tcp://clickhouse:9000 --up=
    deploy:
      restart_policy:
        condition: on-failure
        delay: 5s

  signoz:
    <<: *common
    image: signoz/signoz:${SIGNOZ_IMAGE_TAG:-v0.108.0}
    command:
      - --config=/root/config/prometheus.yml
    environment:
      - SIGNOZ_ALERTMANAGER_PROVIDER=signoz
      - SIGNOZ_TELEMETRYSTORE_CLICKHOUSE_DSN=tcp://clickhouse:9000
      - SIGNOZ_SQLSTORE_SQLITE_PATH=/var/lib/signoz/signoz.db
      - DASHBOARDS_PATH=/root/config/dashboards
      - STORAGE=clickhouse
      - GODEBUG=netdns=go
      - TELEMETRY_ENABLED=true
      - DEPLOYMENT_TYPE=docker-swarm
      - DOT_METRICS_ENABLED=true
    configs:
      - source: signoz-prometheus-config
        target: /root/config/prometheus.yml
    volumes:
      - sqlite:/var/lib/signoz/
      - ./dashboards:/root/config/dashboards:ro
    deploy:
      <<: *deploy_base
      replicas: 1
      labels:
        - 'caddy=signoz.fluxer.app'
        - 'caddy.reverse_proxy={{upstreams 8080}}'
        - 'caddy.header.Strict-Transport-Security="max-age=31536000; includeSubDomains; preload"'
        - 'caddy.header.X-Xss-Protection="1; mode=block"'
        - 'caddy.header.X-Content-Type-Options=nosniff'
        - 'caddy.header.Referrer-Policy=strict-origin-when-cross-origin'
        - 'caddy.header.X-Frame-Options=DENY'
    healthcheck:
      test:
        - CMD
        - wget
        - --spider
        - -q
        - localhost:8080/api/v1/health
      interval: 30s
      timeout: 5s
      retries: 3

  otel-collector:
    <<: *common
    image: signoz/signoz-otel-collector:${OTELCOL_TAG:-v0.129.12}
    command:
      - --config=/etc/otel-collector-config.yaml
      - --manager-config=/etc/manager-config.yaml
      - --copy-path=/var/tmp/collector-config.yaml
      - --feature-gates=-pkg.translator.prometheus.NormalizeName
    configs:
      - source: otel-collector-config
        target: /etc/otel-collector-config.yaml
      - source: otel-manager-config
        target: /etc/manager-config.yaml
    environment:
      - OTEL_RESOURCE_ATTRIBUTES=host.name={{.Node.Hostname}},os.type={{.Node.Platform.OS}}
      - LOW_CARDINAL_EXCEPTION_GROUPING=false
    ports:
      - '4317:4317'
      - '4318:4318'
    deploy:
      <<: *deploy_base
      replicas: 3

networks:
  fluxer-shared:
    external: true

volumes:
  clickhouse:
    driver: local
  sqlite:
    driver: local
  zookeeper-1:
    driver: local

configs:
  clickhouse-config:
    file: ./conf/clickhouse/config.xml
  clickhouse-users:
    file: ./conf/clickhouse/users.xml
  clickhouse-custom-function:
    file: ./conf/clickhouse/custom-function.xml
  clickhouse-cluster:
    file: ./conf/clickhouse/cluster.xml
  signoz-prometheus-config:
    file: ./conf/signoz/prometheus.yml
  otel-collector-config:
    file: ./conf/signoz/otel-collector-config.yaml
  otel-manager-config:
    file: ./conf/signoz/otel-collector-opamp-config.yaml