Docker is an open-source project that makes creating and managing Linux containers really easy. Containers are like extremely lightweight VMs – they allow code to run in isolation from other containers but safely share the machine’s resources, all without the overhead of a hypervisor.
Docker containers can boot extremely fast (in milliseconds!) which gives you unprecedented flexibility in managing load across your cluster. For example, instead of running chef on each of your VMs, it’s faster and more reliable to have your build system create a container and launch it on the appropriate number of CoreOS hosts. This guide will show you how to launch a container, install some software on it, commit that container, and optionally launch it on another CoreOS machine. Before starting, make sure you've got at least one CoreOS machine up and running — try it on Amazon EC2 or locally with Vagrant.
Docker has a straightforward CLI that allows you to do almost everything you could want to a container. All of these commands use the image id (ex. be29975e0098), the image name (ex. myusername/webapp) and the container id (ex. 72d468f455ea) interchangeably depending on the operation you are trying to do. This is confusing at first, so pay special attention to what you're using.
Launching a container is simple as
docker run + the image name you would like to run + the command to run within the container. If the image doesn't exist on your local machine, docker will attempt to fetch it from the public image registry. Later we'll explore how to use docker with a private registry. It's important to note that containers are designed to stop once the command executed within them has exited. For example, if you ran
/bin/echo hello world as your command, the container will start, print hello world and then stop:
docker run ubuntu /bin/echo hello world
Let's launch an Ubuntu container and install Apache inside of it using the bash prompt:
docker run -t -i ubuntu /bin/bash
-i flags allocate a pseudo-tty and keep stdin open even if not attached. This will allow you to use the container like a traditional VM as long as the bash prompt is running. Install Apache with
apt-get update && apt-get install apache2. You're probably wondering what address you can connect to in order to test that Apache was correctly installed...we'll get to that after we commit the container.
After that completes, we need to
commit these changes to our container with the container ID and the image name.
To find the container ID, open another shell (so the container is still running) and read the ID using
The image name is in the format of
username/name. We're going to use
coreos as our username in this example but you should sign up for a Docker.IO user account and use that instead.
It's important to note that you can commit using any username and image name locally, but to push an image to the public registry, the username must be a valid Docker.IO user account.
Commit the container with the container ID, your username, and the name
docker commit 72d468f455ea coreos/apache
The overlay filesystem works similar to git: our image now builds off of the
ubuntu base and adds another layer with Apache on top. These layers get cached separately so that you won't have to pull down the ubuntu base more than once.
Now we have our Ubuntu container with Apache running in one shell and an image of that container sitting on disk. Let's launch a new container based on that image but set it up to keep running indefinitely. The basic syntax looks like this, but we need to configure a few additional options that we'll fill in as we go:
docker run [options] [image] [process]
The first step is to tell docker that we want to run our
docker run [options] coreos/apache [process]
When running docker containers manually, the most important option is to run the container in detached mode with the
-d flag. This will output the container ID to show that the command was successful, but nothing else. At any time you can run
docker ps in the other shell to view a list of the running containers. Our command now looks like:
docker run -d coreos/apache [process]
Do not run containers with detached mode inside of systemd unit files. Detached mode prevents your init system, in our case systemd, from monitoring the process that owns the container because detached mode forks it into the background. To prevent this issue, just omit the
-d flag if you aren't running something manually.
We need to run the apache process in the foreground, since our container will stop when the process specified in the
docker run command stops. We can do this with a flag
-D when starting the apache2 process:
/usr/sbin/apache2ctl -D FOREGROUND
Let's add that to our command:
docker run -d coreos/apache /usr/sbin/apache2ctl -D FOREGROUND
While the sections above explained how to run a container when configuring it, for a production setup, you should not manually start and babysit containers.
Instead, create a systemd unit file to make systemd keep that container running. See the Getting Started with systemd for details.
The default apache install will be running on port 80. To give our container access to traffic over port 80, we use the
-p flag and specify the port on the host that maps to the port inside the container. In our case we want 80 for each, so we include
-p 80:80 in our command:
docker run -d -p 80:80 coreos/apache /usr/sbin/apache2ctl -D FOREGROUND
You can now run this command on your CoreOS host to create the container. You should see the default apache webpage when you load either
localhost:80 or the IP of your remote server. Be sure that any firewall or EC2 Security Group allows traffic to port 80.
Earlier we downloaded the ubuntu image remotely from the docker public registry because it didn't exist on our local machine. We can also push local images to the public registry (or a private registry) very easily with the
docker push coreos/apache
To push to a private repository the syntax is very similar. First, we must prefix our image with the host running our private registry instead of our username. List images by running
docker images and insert the correct ID into the
docker tag f455ea72d468 registry.example.com:5000/apache
After tagging, the image needs to be pushed to the registry:
docker push registry.example.com:5000/apache
Once the image is done uploading, you should be able to start the exact same container on a different CoreOS host by running:
docker run -d -p 80:80 registry.example.com:5000/apache /usr/sbin/apache2ctl -D FOREGROUND