docker-compose ps does not show running services

Question

I have a docker-compose stack launched on a remote machine, through gitlab CI/CD (a runner connects to the docker engine on the remote machine and performs the deploy with docker-compose up -d).

When I connect to that machine from my laptop, using eval docker-machine env REMOTE_ADDRESS, I can see the docker processes running (with docker ps), while the services stack appears to be empty (docker-compose ps).

I am not able to use docker-compose down to stop the stack, and trying docker-compose up -d gives me the error

ERROR: for feamp_postgres  Cannot create container for service postgres: Conflict. The container name "/feamp_postgres" is already in use by container "40586885...". You have to remove (or rename) that container to be able to reuse that name.

The reverse is also true, I can start the stack from my local laptop (using docker-machine), but then the CI/CD pipeline fails when trying to execute docker-compose up -d with the same error.

This happens using the latest versions of docker and docker-compose, both on the laptop (OSX) and on the runner (ubuntu 18.04).

In other circumstances (~10 other projects) this has worked smoothly.

This is the docker-compose.yml file I am using.

version: "3.7"

services:

  web:
    container_name: feamp_web
    restart: always
    image: guglielmo/fpa/feamp:latest
    expose:
      - "8000"
    links:
      - postgres:postgres
      - redis:redis
    environment:
      - ...
    volumes:
      - public:/app/public
      - data:/app/data
      - uwsgi_spooler:/var/lib/uwsgi
      - weblogs:/var/log
    command: /usr/local/bin/uwsgi --socket=:8000 ...

  nginx:
    container_name: feamp_nginx
    restart: always
    ...

  postgres:
    container_name: feamp_postgres
    restart: always
    image: postgres:11-alpine
    ...

  redis:
    container_name: feamp_redis
    restart: always
    image: redis:latest
    ...

volumes:
    ...

networks:
  default:
    external:
      name: webproxy

Normally I can up the stack from my local laptop and manage it from the CI/CD pipeline on gitlab, or vice-versa.

This diagram should help visualise the situation.

                        +-----------------+                          
                        |                 |                          
                        |  Remote server  |                          
                        |                 |                          
                        +----|--------|---+                          
                             |        |                              
                             |        |                              
            docker-compose ps|        |docker-compose up -d          
                             |        |                              
                             |        |                              
+-------------------+        |        |        +--------------------+
|                   |        |        |        |                    |
|  Docker client 1  ---------+        +---------  Docker client 2   |
|                   |                          |                    |
+-------------------+                          +--------------------+

Connection to the remote server's docker engine are executed through docker-machine.

Answer 1

It appears that specifying the project name when invoking docker-compose commands, solves the issue.

This can be done using the -p parameter in the command line or the COMPOSE_PROJECT_NAME environment variable, on both clients.

For some reasons, this was not needed in previous projects. It may be a change in docker (I changed from 18 to 19), or something else, I still do not know the details.