Table of contents
- Why Docker
- Writing a Dockerfile
- Deploying to AWS
- A simple deploy script
Why would you want to go through the hassle of deploying your prototypes in the first place? I think there are a couple of things that makes it worth your effort.
First off, I personally get way more motivated when I know that the thing I’m working on is actually live somewhere on the web, even if it’s behind an obscure IP address. It’s much more fun to add a feature or fix a bug when you know you can push it live easily.
Secondly, in the best case scenario your prototype will grow into a proper product at some point and you’ll have to deploy it anyway. By deploying early, even if all it does is print ‘‘Hello World’’, you’ll catch issues along the way which will allow you to incrementally improve your product in a way that’s still deployable in the end.
Finally I hope to show you that if you’re willing to compromise on some of the aspects of deploying your service (such as availability, auto scaling etc.) then it really isn’t a lot of work. In fact I’ll show you a single shell script you can run that gets the job done.
It was a fun experience helping my brother deploy his service as the requirements were so different from what I usually have to deal with when automating our deployments at Famly. I worry about the availability of the service, configuring auto-scaling, ensuring the logs are persisted and queryable, and things like that. However, in this case we didn’t really care about any of these things – Instead our main focus was simplicity and ease of automation.
Simplicity in this case meant having as few moving parts as possible and keeping the number of new concepts my brother had to learn to a minimum, while ensuring that the ones he did have to learn would be useful to him in other contexts as well. I decided that the simplest solution would be to use docker and docker-machine.
From the perspective of using tools that might be useful to know in other contexts I believe that Docker is an obvious choice; if you know how to containerize your application then you’ll be able to run it almost anywhere (contrast that to learning how to set up your application using Elastic Beanstalk for example). Given that you can containerize almost anything this means that if you’re familiar with Docker you’ll be able to run almost anything almost anywhere 😉
The ability to iterate quickly on the Dockerfile and run it locally before trying to deploy it was also a key factor; once it runs locally it you should be able to deploy it easily as well.
Writing a Dockerfile
If you don’t want to write a Dockerfile for one of your own projects but still want to try deploying something to AWS then feel free to use the small example I’ve created.
Deploying to AWS
At this point I assume you have a Docker image that you’ve been able to run locally, so now it’s time to deploy it. 🚀
Provisioning a machine
docker-machine is a tool that makes it possible to run
docker commands on
your machine as you’re used to, but in reality they will be running on a remote
Before you can deploy anything you need to have somewhere to deploy it to; you
need to rent a server. Normally with AWS you’d have to launch an EC2
instance, give it the right roles etc. This is where
docker-machine comes in
handy. The following command will provision a machine and open port 80 so it can
accept HTTP traffic.
docker-machine create \ --amazonec2-open-port 80 \ --amazonec2-region eu-central-1 \ --amazonec2-instance-type t2.micro \ --driver amazonec2 \ <YOUR_MACHINE_NAME>
The instance type is set to
t2.micro which is one of the cheapest server you
can buy (See the full list of instance types here for the
You can read the documentation for each of the command line arguments here.
If you have installed Docker for Mac then you might have an outdated version
docker-machine --version should be
0.12.2 or higher. If you’ve installed
docker-machine through homebrew you’ll need to modify your shell
Note: As of Docker for Mac 17.09.0-ce-mac35 (19611) it ships with version
docker-machine so you don’t have to install it separately.
Deploying the service
Now that we have a machine we can deploy our service to it. To run your
commands on the remote Docker machine you first have to tell
docker to execute
commands in that context. You do that by using the following command.
eval $(docker-machine env <YOUR_MACHINE_NAME>)
docker commands in your shell session will be executed on the remote
docker machine. If you close your terminal you’ll have to run the command
above again in case you want to speak with the remote docker machine.
If you’re curious how this works then simply run
docker-machine env <YOUR_MACHINE_NAME> without the
eval prefix. You’ll
see that all it does it set some environment variables that Docker reads
in order to know what Docker machine to talk to.
Before you can run the image on the machine you first have to build the image on it. This will basically transfer all of the files in your current folder unto the machine and then build the image there. If you’re using my small example then you can build the image with the following command
docker build --tag deploying-prototypes:local .
Finally start a Docker container based on the image.
docker run \ --detach \ --publish 80:80 \ deploying-prototypes:local
See that it works 🎉
echo http://$(docker-machine ip <YOUR_MACHINE_NAME>)
Once you’re done experimenting you might want to stop the machine so you won’t be billed if you’re not using it.
docker-machine stop <YOUR_MACHINE_NAME>
A simple deploy script
In the section above you deployed the service manually by writing a set of
commands in the shell. Once you’ve done that a couple of times you’ll get tired
of it, so here’s a deploy script that I wrote that automates it for you. It
assumes that your Docker image listens for HTTP traffic on port 80 and that port
80 is exposed on the machine (which it is if you used the
command above to create it).
Download the script and make it executable.
curl https://gist.githubusercontent.com/mads-hartmann/415cba506a538f35a992598c9221432d/raw/98260522a6358feb6c4b70ad503c2e6bbe9b5ce8/prototype-deploy.sh > prototype-deploy.sh chmod +x prototype-deploy.sh
Now you can deploy any Docker service to any Docker machine like this
(assuming your current folder contains the script and a
./prototype-deploy.sh <YOUR_MACHINE_NAME> <IMAGE_NAME> <CONTAINER_NAME>
<CONTAINER_NAME> is completely up to you.
Have fun deploying your next prototype, and let me know in the comments below if you found it useful or have any questions 🙌