Some Docker usage

A project log for Learning Docker

Docker is a popular containerization (OS-level virtualization) tool. I'm learning how to use it. You can follow along if you want.

pointyointmentPointyOintment 04/01/2019 at 05:500 Comments

This is based on a demo I did in class recently, but expanded to cover more advanced topics that I had too little time to talk about then.

Part 1: Basics

Hello World, etc.

I assume you have already installed Docker and have a command line such as Docker Quickstart Terminal. To make sure it's working, type

$ docker run hello-world

That command will tell Docker to run a container based on an image called hello-world. If you haven't done this before, you won't have that image, so Docker will have to download it first. Then the container will run. It contains a program that just prints out a message explaining the process of running the container and getting the output from it back to your terminal. Once this is done, the container shuts down, but the image remains on your computer in case you want to run a container based on it again.

To see that you still have that image, you can type

 docker image ls

which will print a list of all images you have on your host. To see that the container is no longer running, you can type

$ docker container ls

which will print a list of all currently running containers (none, now that the hello-world container has shut down). That container still exists, though—you can use this command to show a list of all containers including those not currently running:

$ docker container ls -a

The hello-world output suggests running the below command:

$ docker container run -it ubuntu /bin/bash

This will start an Ubuntu Linux container, and give you an interactive session. -i makes sure STDIN stays open, and -t connects your terminal to a terminal in the container. You also have to specify what shell executable to run once the container is up (if a shell is what you want).

In my demo, I replaced ubuntu with alpine, to use Alpine Linux, because it would download faster. It's a lightweight Linux distribution, which doesn't come with everything you might expect, but it was sufficient for my purposes. It's a popular base for Docker projects, too.

Anyway, when you're done playing with it, you can type exit to disconnect from the container, at which point it will shut down. (The program you told Docker to run when you started the container—bash—has exited, so Docker considers it time to shut down the container.)

You can also run programs other than shells, like so (noting the absence of -it):

$ docker container run ubuntu ls -l

which will list the contents of the root directory with details, and then the container will shut down. Even without -i or -t, the output will be routed by Docker from the virtual terminal in the container to the terminal you're typing in.

Isolation between containers, and a bit of container management

You can run multiple containers based on the same image, and they will automatically be isolated from each other. Each has its own filesystem, and they cannot communicate with each other unless you tell Docker to let them. For example, if we start a container:

 docker run -it alpine /bin/bash

and create a file:

$ echo "secret text" > secret.txt

we can then, still in the same container, work with that file. But once we exit the container and start a new one from the same image:

$ exit
$ docker run alpine ls

we cannot see that file. It doesn't exist in this new container. But we can still get back into the container in which we created the file. First we need to know the shut-down container's ID:

$ docker container ls -a

Find the container in the list that matches the command you started it with and the time when you started it. Note the ID listed. You only need as much of the beginning of the ID as is necessary to distinguish it from other IDs. (The real IDs are actually much longer than what's shown, so even if you type in the full ID shown by the above command, you're still using the shortcut.)

Then start that existing container:

$ docker container start <container ID>

OK, now the container is running again. But it's running in the background, and we have no interactive session. So how do we tell it what to do? We can use the exec command:

$ docker container exec <container ID> cat secret.txt
secret text

Now that the container is running in the background, it will keep running even after commands we give it have finished. So, to shut it down, we need to use stop:

$ docker container stop <container ID>

We can also remove containers we no longer need:

$ docker container rm <container ID>

(and similarly for images).

Part 2: Trying to do something slightly useful

Now we'll try to make our own container with custom contents. The contents will be a program called FIGlet that makes ASCII art banner text, as well as a custom Python script that both runs FIGlet and returns the hostname of the container (which Docker sets to the container ID). To do this, we'll need to start with a stock Python container. This way of building custom containers by starting with off-the-shelf containers is pretty common in the world of Docker.

Here's how you make a custom Docker image:

  1. Prepare all of the files you want to put in it (executables, configuration files, etc.) by putting them in a directory on your host
  2. Make a text file with the name Dockerfile (no extension) that specifies how the container is to be built and run
  3. Tell Docker to build it

Here's the Dockerfile we'll use for this example—read the comments to see what everything means:

# Use official Alpine Linux-based Python 3 image as parent image
FROM python:3-alpine

# Update package repository
RUN apk update

# Install FIGlet
RUN apk add figlet

# Set the working directory to /app

# Copy the requirements file into the container
COPY requirements.txt ./

# Install any needed packages specified in requirements.txt
RUN pip install --no-cache-dir --trusted-host -r requirements.txt

# Copy anything else in
COPY . .

# Define environment variable
ENV NAME a-container

# Run when the container launches
CMD ["python", ""]

Put it in a work directory on your host. In the same directory, put a Python file called <a target="_blank" rel="noopener noreferrer" href=""></a> and a text file called requirements.txt. requirements.txt is actually empty right now; it would list the names of Python packages to be installed by pip (Python's package manager) during the Docker container build process, if we needed any. <a target="_blank" rel="noopener noreferrer" href=""></a> contains this code:

import os, socket, subprocess

if __name__ == "__main__":
    name = os.getenv("NAME", "none")
    hostname = socket.gethostname()
    print("Hello from " + name + "/" + hostname)["echo", "Python"]) 

(You might notice that FIGlet is not being called from Python. Wait and see.) Once those files are in place, navigate to your work directory in your Docker terminal, and run this command:

$ docker build -t python-demo:v0.1 .

Don't forget the dot at the end! That tells Docker to build from the contents of the current directory. If you don't want to cd to your work directory before building, you can just put the path in place of the dot. -t is for "tag", which comes in two parts: python-demo is the repository name, and v0.1 is the tag proper (which is usually used for a version number).

Docker will take a while to build your image, but it provides nice verbose output about what steps it's taking. When it's done, you can check that your image exists using docker image ls

$ docker run python-demo:v0.1

It should output two lines of text, one just Python and one that says Hello from <hostname/container ID>, and then shut down.

Now we'll have it use FIGlet. Change's last line to:["figlet", "Python"]) 

Now we have to rebuild the image:

$ docker build -t python-demo:v0.2 .

Note that the version number is now 0.2. Once it's done, if you run docker image ls, you should see both versions. You can still run v0.1 and it will produce the same output as before. But running v0.2:

$ docker run python-demo:v0.2

should produce a nice "Python" ASCII art banner instead of just text, in addition to the same Hello from <hostname/container ID> line. (The outputs appear out of order because the external program, echo or figlet, runs in parallel to the Python script.)

You can also get an interactive session with a container based on this image:

$ docker run -it python-demo:v0.2 /bin/bash

and then you can do whatever you want from the command line within the container, including using FIGlet to make other banners. Don't forget to clean up containers you won't use again by shutting them down and deleting them.