bpp-deploy/docker-compose.workers.yml at main · iplweb/bpp-deploy · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
x-logging: &default-logging
  driver: "local"
  options:
    max-size: "${LOG_MAX_SIZE:-150m}"
    max-file: "${LOG_MAX_FILE:-5}"

services:

  workerserver-general:
    image: iplweb/bpp_workerserver:${DOCKER_VERSION:-latest}
    restart: always
    logging: *default-logging
    env_file: ${BPP_CONFIGS_DIR}/.env
    volumes:
      - staticfiles:/staticroot
      - media:/mediaroot
    depends_on:
      # Musi czekać na appserver, gdyż appserver aktualizuje ewentualnie baze danych.
      # Czy nawet w ogóle ją buduje...
      appserver:
        condition: service_healthy
    deploy:
      resources:
        limits:
          memory: ${WORKER_GENERAL_MEM_LIMIT:-1g}
          cpus: "${WORKER_GENERAL_CPU_LIMIT:-2.0}"
    labels:
      ofelia.enabled: "true"
      # Nocny restart zbija skumulowany memory leak Celery workera (Python long-running).
      # kill 1 -> SIGTERM do PID 1 -> graceful shutdown -> restart:always wskrzesza.
      ofelia.job-exec.restart_self.schedule: "0 5 5 * * *"
      ofelia.job-exec.restart_self.command: "kill 1"

  workerserver-denorm:
    image: iplweb/bpp_workerserver:${DOCKER_VERSION:-latest}
    restart: always
    logging: *default-logging
    env_file: ${BPP_CONFIGS_DIR}/.env
    environment:
      CELERY_QUEUE: "denorm"
    volumes:
      - staticfiles:/staticroot
      - media:/mediaroot
    depends_on:
      # Musi czekać na appserver, gdyż appserver aktualizuje ewentualnie baze danych.
      # Czy nawet w ogóle ją buduje...
      appserver:
        condition: service_healthy
    deploy:
      resources:
        limits:
          memory: ${WORKER_DENORM_MEM_LIMIT:-1g}
          cpus: "${WORKER_DENORM_CPU_LIMIT:-1.0}"
    labels:
      ofelia.enabled: "true"
      ofelia.job-exec.restart_self.schedule: "0 10 5 * * *"
      ofelia.job-exec.restart_self.command: "kill 1"

  denorm-queue:
    # run only SINGLE service, this is only a mechanism to pass
    # PostgreSQL LISTEN to Celery queue
    image: iplweb/bpp_denorm_queue:${DOCKER_VERSION:-latest}
    logging: *default-logging
    env_file: ${BPP_CONFIGS_DIR}/.env
    # entrypoint: ["python", "src/manage.py", "denorm_queue"]
    # healthcheck:
    #   test: ["CMD-SHELL", "pgrep -f 'denorm_queue' > /dev/null"]
    #   start_period: 10s
    #   interval: 30s
    #   timeout: 5s
    #   retries: 3
    restart: always
    depends_on:
      workerserver-denorm:
        condition: service_healthy
    deploy:
      resources:
        limits:
          memory: ${DENORM_QUEUE_MEM_LIMIT:-320m}
          cpus: "${DENORM_QUEUE_CPU_LIMIT:-1.0}"
    labels:
      ofelia.enabled: "true"
      # Restart najpozniej w staggered secie - po wszystkich pozostalych, zeby NOTIFY
      # ze swiezo podniesionych worker-denormow mialy do czego trafic.
      ofelia.job-exec.restart_self.schedule: "0 25 5 * * *"
      ofelia.job-exec.restart_self.command: "kill 1"

  workerserver-status:
    image: iplweb/bpp_workerserver:${DOCKER_VERSION:-latest}
    logging: *default-logging
    env_file: ${BPP_CONFIGS_DIR}/.env
    entrypoint: celery -A django_bpp.celery_tasks status
    depends_on:
      workerserver-general:
        condition: service_healthy
      workerserver-denorm:
        condition: service_healthy
    profiles: ['manual']

  celerybeat:
    image: iplweb/bpp_beatserver:${DOCKER_VERSION:-latest}
    logging: *default-logging
    env_file: ${BPP_CONFIGS_DIR}/.env
    restart: always
    volumes:
      - staticfiles:/staticroot
      - media:/mediaroot
    depends_on:
      redis:
        condition: service_healthy
      rabbitmq:
        condition: service_healthy
      appserver:
        condition: service_started
    deploy:
      resources:
        limits:
          memory: ${CELERYBEAT_MEM_LIMIT:-320m}
          cpus: "${CELERYBEAT_CPU_LIMIT:-0.25}"
    labels:
      ofelia.enabled: "true"
      ofelia.job-exec.restart_self.schedule: "0 20 5 * * *"
      ofelia.job-exec.restart_self.command: "kill 1"

  flower:
    image: mher/flower:2.0.1
    restart: always
    logging: *default-logging
    environment:
      - CELERY_BROKER_URL=amqp://${DJANGO_BPP_RABBITMQ_USER:-bpp}:${DJANGO_BPP_RABBITMQ_PASS:-bpp}@rabbitmq:${DJANGO_BPP_RABBITMQ_PORT:-5672}//
      - FLOWER_PORT=5555
      - FLOWER_URL_PREFIX=/flower
    depends_on:
      rabbitmq:
        condition: service_healthy
    expose:
      - 5555
    deploy:
      resources:
        limits:
          memory: ${FLOWER_MEM_LIMIT:-768m}
          cpus: "${FLOWER_CPU_LIMIT:-0.5}"
    labels:
      ofelia.enabled: "true"
      # Flower akumuluje historie zadan Celery; nocny restart zbija konsumpcje pamieci.
      # mher/flower 2.0.1 opiera sie na python:3.11-slim - ma /bin/kill z procps,
      # ale Docker exec uruchamia go jako root przez PID namespace kontenera,
      # wiec "kill 1" wysyla SIGTERM do PID 1 bezposrednio (bez shell-a).
      ofelia.job-exec.restart_self.schedule: "0 15 5 * * *"
      ofelia.job-exec.restart_self.command: "kill 1"