Docker is a lightweight container for applications – think of a Docker as an app in a box, except that the box in this case isn’t an entire VM, but the bare necessities required to run a process. Consequently, you can run many Dockers in a VM. In essence, Docker replaces installation steps for a particular app. Rather than having to execute a series of steps to get, say, MongoDB running, you can simply fire up a Mongo Docker image.

Docker images can be created from a Dockerfile, which is similar to a Vagrantfile or even a build script – it’s a prescription for how to assemble an image. You don’t need to have a Dockerfile to create a Docker image, however, creating one makes image creation repeatable. It also provides a means for others to verify an image.

There are a few key instructions you should be aware of when creating Dockerfiles – mainly, FROM, RUN, EXPOSE, and CMD. To demonstrate how easy this process is, I’m going to create a Dockerfile that runs Amazon’s DynamoDB Local.

DynamoDB Local

DynamoDB Local is a simple Java application that emulates AWS DynamoDB. The benefit of using DynamoDB Local is that you can iterate quickly without using bandwidth against the real DynamoDB and you’ll save a few coins in the process.

Running DynamoDB Local isn’t terribly difficult; in fact, provided you have Java installed, it’s as easy as:

1

$ java -Djava.library.path=./DynamoDBLocal_lib -jar DynamoDBLocal.jar

Of course, if you aren’t developing a Java application and don’t have Java installed, you would need to, of course, install Java. Then you’d need to download the DynamoDB Local package and install it. And then you’d need to run it.

Alternatively, you could just pull a Docker image, run it, and get back to work.

Creating a Dockerfile

Creating a Dockerfile is simple. Fire up your favorite editor and follow along.

The first required element, FROM, indicates the base image or parent from which a docker image is built upon. In many cases, this’ll be something like ubuntu or centos, for example. In the case of an image for DynamoDB Local, I’m going to base it off another image that already has Oracle’s Java 8 installed. That image is based upon Ubuntu. Note, basing your FROM off of another image implies the image is available in an accessible Docker index. You can run your own local or remote indexes or use Docker’s public index.

Consequently, the first two lines of my Dockerfile are:

1
2

FROM aglover/java8-pier
MAINTAINER Andy Glover "ajglover@gmail.com"

Have a look at my java8-pier project and you’ll see its Dockerfile has the line FROM ubuntu. The MAINTAINER line is self evident.

The 2nd most important instruction of a Dockerfile is RUN. Think of RUN as bash commands required to set up your Docker image. In the case of setting up DynamoDB Local, there are a few things I would like done on the image. First and foremost, I’d like to update base aspects of the underlying Ubuntu OS via an apt-get update. Then I’d like to download the DynamoDB Local archive, however, I’d like to use wget, which isn’t available on base Ubuntu installs, however. Consequently, I’ll install wget while I’m at it.

1
2

RUN apt-get update -y
RUN apt-get install wget -y

Next, I’m going to wget the latest version of DynamoDB Local via an AWS URL that points to the latest version (which obviously changes); thus, the -O flag forces the downloaded file to the generic name of dynamo.tar.gz (rather than something like dynamodb_local_2014-04-24.tar.gz where the date can change depending on when AWS releases an update). Finally, after the download completes, the file is moved into a directory dubbed dynamodb_local.

1
2

RUN wget http://dynamodb-local.s3-website-us-west-2.amazonaws.com/dynamodb_local_latest -O dynamo.tar.gz
RUN tar xvzf dynamo.tar.gz && mv dynamodb_local_* dynamodb_local

Docker images, by default, don’t expose any ports through the host on which they are running. You must expose desired ports via the EXPOSE command. In my case, DynamoDB Local defaults to port 8000; accordingly, I’ll specify in my Dockerfile that I wish this port to be open:

1

EXPOSE 8000

Finally, as I wish to run a service via my Docker image, I need to fire it up! DynamoDB Local is simply a Java process that requires, at a minimum, two parameters. If I were to run DynamoDB Local manually, the corresponding command would be:

1

$ java -Djava.library.path=./dynamodb_local/DynamoDBLocal_lib -jar ./dynamodb_local/DynamoDBLocal.jar

Consequently, to execute this command in a Dockerfile I’ll need to use the CMD instruction (which is probably the most important instruction for creating Dockers!). This instruction takes an array of values – logically, just take the corresponding manual command and tokenize it by a space and you’ve got your CMD:

1

CMD ["java", "-Djava.library.path=./dynamodb_local/DynamoDBLocal_lib", "-jar", "./dynamodb_local/DynamoDBLocal.jar"]

That’s it – 8 lines and you’ve got a Dockerfile that’ll yield Docker image running DynamoDB Local as a service.

Creating images from Dockerfiles

With a Dockerfile I can now create an image via the build command. Thus, in a terminal window, I’ll change directories to where I’ve saved my Dockerfile and run the following:

1

$ docker build -t aglover/dynamodb-pier .

aglover/dynamodb-pier is the desired name of my image. After you run this command, you will see a whole lot of commands executed, including the ones specified in your Dockerfile. Once things finish successfully, you should be able to see the resultant image via the images command.

1
2
3
4
5
6
7
8

$ docker images
REPOSITORY              TAG                 IMAGE ID            CREATED             VIRTUAL SIZE
aglover/dynamodb-pier   latest              aa51ccc5dc3f        14 seconds ago      1.089 GB
aglover/java8-pier      latest              cb8436bb816a        4 weeks ago         1.026 GB
base                    latest              b750fe79269d        13 months ago       175.3 MB
base                    ubuntu-12.10        b750fe79269d        13 months ago       175.3 MB
base                    ubuntu-quantal      b750fe79269d        13 months ago       175.3 MB
base                    ubuntu-quantl       b750fe79269d        13 months ago       175.3 MB

I can run my newly minted Docker image via its ID, which, if you look closely in the listing above, is aa51ccc5dc3f.

1
2
3

$ docker run aa51ccc5dc3f
2014-05-05 17:04:14.037:INFO:oejs.Server:jetty-8.1.12.v20130726
2014-05-05 17:04:14.119:INFO:oejs.AbstractConnector:Started SelectChannelConnector@0.0.0.0:8000

Note, by default, Docker will run the image in the foreground; accordingly, you can see things are working via the output coming from the DynamoDB Local instance running.

You can run a Docker image as a daemon via the -d flag:

1
2

$ docker run -d aa51ccc5dc3f
7be2708d94eedaa82432d239659ddb696d66004516174a6e2f79f4ec465eb9fc

You can now see what Docker images are running via the ps command.

1
2
3

$ docker ps
CONTAINER ID        IMAGE                          COMMAND                CREATED             STATUS              PORTS               NAMES
7be2708d94ee        aglover/dynamodb-pier:latest   java -Djava.library.   17 seconds ago      Up 16 seconds       8000/tcp            compassionate_bardeen

Finally, you can stop an image via the stop command. You must provide the ID of the image, which you can see via a ps.

1

$ docker stop 7be2708d94ee

Docker has a public repository can you publish to, provided you have an account. Ultimately, publishing is done via the push command. For example, I’ve published my DynamoDB Local image via:

1

$ docker push aglover/dynamodb-pier

Once an image has been published, it can correspondingly be downloaded via the pull command:

1

$ docker pull aglover/dynamodb-pier

Think of pulling as simply downloading and registering a Docker. To use a ‘pull’ed Docker, you must still execute the run command.

Docker makes it super easy to distribute pre-packaged applications – rather than installing various binaries on different operating systems (like MongoDB on OSX for development and a different MongoDB binary on Ubuntu for production), you can use the same package across environments. Dockerfiles make creating Dockers repeatable. And hopefully I’ve shown you that crafting a Dockerfile isn’t terribly difficult. Go forth and Docker.