Demonstrating the difference between detached and not detached containers

Question

I have a very simple dockerfile:

FROM alpine:3.14 AS build

LABEL version = "1.0"
LABEL description = "Our first Docker file!"

and a very simple docker-compose.yml file:

version: "3.9"

# Define services
services:
    container1:
        container_name: first_container
        build:
            context: .
            dockerfile: dockerfile

I am running the docker-compose file with docker-compose up and, alternatively, docker-compose up -d to demonstrate that the container is still running in the detached configuration.

In the first case, I expect the container to run then close. It does. Running docker-compose down then stops and removes all the containers (first_container) as confirmed with docker-compose ps.

$ docker-compose up
Creating network "comdockerdevenvironmentscode_default" with the default driver
Creating first_container ... done
Attaching to first_container
first_container exited with code 0

$ docker-compose ps
     Name         Command   State    Ports
------------------------------------------
first_container   /bin/sh   Exit 0        

$ docker-compose down
Removing first_container ... done
Removing network comdockerdevenvironmentscode_default

$ docker-compose ps
Name   Command   State   Ports
------------------------------

Now, I rerun docker-compose up again, adding the detached command. Now, the output says it creates the container, but doesn't close it. Great! I should see a status change when running docker-compose ps, but it still seems to have exited.

$ docker-compose up -d
Creating network "comdockerdevenvironmentscode_default" with the default driver
Creating first_container ... done

$ docker-compose ps
     Name         Command   State    Ports
------------------------------------------
first_container   /bin/sh   Exit 0        

Why is this? How can I show that I can attach to the running detached container, but not to the closed container in the other case?

Answer 1

The alpine container, absent an explict configuration to the contrary, starts a shell (/bin/sh) by default. The shell exits immediately if stdin isn't open. Regardless of whether or not you run docker-compose up with -d, in all cases stdin is closed, so your container exits.

If you want the container to run a shell without exiting, then you need to arrange to stdin to remain connected. To get a prompt, you'll also need to have Docker allocate a tty. That means:

version: "3.9"

services:
    container1:
        image: docker.io/alpine:latest
        stdin_open: true
        tty: true

Running a container as in the above example means that you can docker attach to it and access the shell prompt.

Alternately, if you'd like the container to run without exiting, just run a command that won't exit:

version: "3.9"

services:
    container1:
        image: docker.io/alpine:latest
        command: [sleep, inf]

This runs the sleep inf command, which runs until canceled. A more typical example would run some sort of persistent service like a web server, a database, etc.