I considered it, seems nice.

Problem is that they recommend a 12 physical cores and 12GB which is a waste for the usual selfhosted lab.

Since I was feeling bad for giving the wrong answer in another comment, I spin up a docker socket proxy and did some test :)

The main points are:

• add DOCKER_HOST variable pointing to your tcp socket
• add CONTAINERS=1 variable in docker-socker-proxy to allow reading containers, otherwise it will fails silently (unless you run glances with -d) with a 403

Here’s a sample compose file, adjust to your needs. Please note that the tcp socket is not exposed outside of admin_net network and that glances does not have access to the docker.sock socket:

    version: '3.3'
    services:
      admin-glances:
        container_name: glances
        restart: always
        ports:
          - '61208:61208'
        environment:
          - GLANCES_OPT=-w
          - DOCKER_HOST=tcp://dockerproxy:2375
        volumes:
          - './glances/glances.conf:/glances/conf/glances.conf'
    #      - '/var/run/docker.sock:/var/run/docker.sock:ro'
        pid: host
        image: 'nicolargo/glances:latest-full'
        networks:
          admin_net:
    
      admin-docker-socket-proxy:
        container_name: dockerproxy
        hostname: dockerproxy
        image: tecnativa/docker-socket-proxy
        environment:
          - CONTAINERS=1
        volumes:
          - '/var/run/docker.sock:/var/run/docker.sock:ro'
    #    ports:
    #      - '2375:2375'
        networks:
          admin_net:
    
    networks:
      admin_net:
        name: admin_net