Docker swarm php-fpm container keeps restarting

Question

I created a basic LAPP stack which runs flawlessly on localhost with docker-compose. When I try to make it run with Swarm on production server (1 manager only, no workers), all services go up and get replicated (1/1) but the php-fpm one which keeps restarting with no apparent error.

docker service ls

ID             NAME                    MODE         REPLICAS   IMAGE                  PORTS
p0rdrdfmso6x   traefik_reverse-proxy   replicated   1/1        traefik:v2.4           *:80->80/tcp, *:443->443/tcp
e6vwlo9iw2ny   my_stack_apache         replicated   1/1        apache:latest          *:8000->80/tcp
qy5yigbcjryr   my_stack_ftp            replicated   1/1        fauria/vsftpd:latest   *:20-21->20-21/tcp, *:22100-22110->22100-22110/tcp
n5f9v6bd2854   my_stack_php            replicated   0/1        php:latest
rcnbq4vnoz1j   my_stack_postgres       replicated   1/1        postgres:9.5.24

If we focus on php-fpm container :

docker service ps my_stack_php

ID             NAME                  IMAGE        NODE                        DESIRED STATE   CURRENT STATE             ERROR     PORTS
j6mp3ka40cyo   my_stack_php.1        php:latest   node.address                Ready           Ready 2 seconds ago
ezztpsjoglwy    \_ my_stack_php.1    php:latest   node.address                Shutdown        Complete 3 seconds ago
gnqjhwpi5y72    \_ my_stack_php.1    php:latest   node.address                Shutdown        Complete 9 seconds ago
0agr3tw0bb9g    \_ my_stack_php.1    php:latest   node.address                Shutdown        Complete 15 seconds ago
9a6wsdp4tqqn    \_ my_stack_php.1    php:latest   node.address                Shutdown        Complete 21 seconds ago

If I look to the logs : docker service logs my_stack_php

[email protected]    | Interactive shell
[email protected]    | Interactive shell
[email protected]    | Interactive shell
[email protected]    |
[email protected]    |
[email protected]    | Interactive shell
[email protected]    |
[email protected]    |

It behaves like a container running a command which would end with succes in a few seconds. Swarm launches then an new container to keep the restart contract. However, my php-fpm Dockerfile provide the -F argument that should keep the process running :

PHP Dockerfile

FROM centos:7.4

# ... all installs from Centos to add PHP 7.2 from Remi Collet repositories

RUN mkdir -p /run/php-fpm
RUN usermod -a -G ftp apache

WORKDIR /var/www/html

EXPOSE 9000

# Run in foreground as root (in container POV)
CMD ["php-fpm", "-R", "-F"]

docker-compose.yaml

version: '3.9'
services:
  postgres:
    image: "postgres:9.5.24"
    environment:
      POSTGRES_DB: /run/secret/postgres_db
      POSTGRES_USER: /run/secret/postgres_user
      POSTGRES_PASSWORD: /run/secret/postgres_password
    volumes:
      - database:/var/lib/postgresql/data
    secrets:
      - postgres_db
      - postgres_user
      - postgres_password
    deploy:
      resources:
        limits:
          cpus: '0.15'
          memory: 128m
    networks:
      - internal

  apache:
    env_file: .env
    image: apache:latest
    build:
      context: ./docker/images/apache2.4
      dockerfile: prod.Dockerfile
    ports:
      - 8000:80
    environment:
      FPM_HOST: php:9000
    volumes:
      - ./docker/logs/apache/:/var/log/httpd/

    networks:
      - traefik-public
      - internal

    deploy:
      labels:
        - "traefik.enable=true"
        - "traefik.http.routers.my_stack.rule=Host(`my-host.com`)"
        - "traefik.http.routers.my_stack.entrypoints=websecure"
        - "traefik.http.routers.my_stack.tls.certresolver=letsencryptresolver"
        - "traefik.http.services.my_stack.loadbalancer.server.port=80"
        - "traefik.port=80"

      resources:
        limits:
          cpus: '0.15'
          memory: 128m
 php:
    env_file: .env
    image: php:latest
#    links:
#      - ftp
#      - apache
    build:
      context: ./docker/images/php
      dockerfile: prod.Dockerfile
#      args:
#        TIMEZONE: 'Europe/Paris'
    volumes:
      - ftp_data:/var/www/ftp:rw


    networks:
      - internal

    deploy:
      resources:
        limits:
          cpus: '0.20'
          memory: 512m

  ftp:
    env_file: .env
    image: "fauria/vsftpd:latest"
    ports:
      - "20:20"
      - "21:21"
      - "22100-22110:22100-22110"
    environment:
      FTP_USER: apache
      FTP_PASS: /run/secret/automation_client_password
      PASV_ADDRESS: 127.0.0.1
      PASV_MIN_PORT: 22100
      PASV_MAX_PORT: 22110
    volumes:
      - ftp_data:/home/vsftpd/apache:rw

    networks:
      - traefik-public
      - internal

    deploy:
      resources:
        limits:
          cpus: '0.15'
          memory: 128m
volumes:
  ftp_data:
  database:

secrets:
  postgres_db:
    external: true
  postgres_user:
    external: true
  postgres_password:
    external: true
  automation_client_password:
    external: true

networks:
  traefik-public:
    external: true
  internal:
    external: false

Anyone got a clue about this? Any helps/tips will be appreciated.

Answer 1

I don't like this kind of answer, but replica's all went live after complete reboot.