DevUA

Seasoned software engineer. I've worked in many challenging projects. For now, I work as a part of Upwork engineering team. As a software engineer I do things in Java. Using Intelij products. A true fan of CI/CD. Truly believe, that nowadays machines cost less, than developers, that's why I'm always willing to automate everything. Because It works. And saves time. And money. Opensource commiter.

Meta


Dockerizing your apps

Ivan UrsulIvan Ursul

Preface

As some of you may already know, from the last autumn-2015 I’m working as a software engineer in startup company called Upwork, in one of teams. From the time I started working there, I realized, that working on freelance basis is completely different from office work, even if you’re working on a good quality freelance job. Why ? At least, because you cant share some knowledge with your coworkers by cuf of coffee in your lunch time. So I decided to convert my blog to some R&D investigation. That’s why I’d like to warn you, that everything here is just an investigation, which I do on my free time, so don’t take it so critically, if you find something, which you don’t agree with, just comment, and let’s discuss. Technologies are changing every day, and we need to follow the trend, don’t we ?

Why containers ?

Before I can proceed with explanation in my head what is the definition of container, let’s return to the near part times, where we did everything manually: the dyno age, where it was ok to assemble your proeject war manually, and deploy it to the server in the same way.
altThen, the new era started, and we got build systems. It was cool, because we now could build and deploy our software using build commands. But it wasn’t the true way it shold be. Why ? Because it was clear for us, developers, how to build our software, but was it clear for devops ? Of course, not, because there was lot’s of build systems, and for each of them devops needed to find aproach. It was not ideal, because developers wrote dozens of deployment instructions.

From the prospective of past times, now I understand, that if developer had to write some instructions for deployment or even scripts, then he could easily follow some standard or specification – a formal contract, which is known for developer and devops engineer: developer knows how to write deployment instuctions and devops know how to deploy this instuctions.

This resulted in so called application containers – a strucute, where your application is running, and it’s easy to deploy it. One of this containers is Docker. Why it’s cool ? Because you can deploy application without knowing what’s inside. Perfect.

What is docker ?

Wiki page describes docker as an “open-source project that automates the deployment of applications inside software containers”. So, it’s a thing called container. Docker containers wrap up a piece of software in a complete filesystem that contains everything it needs to run: code, runtime, system tools, system libraries – anything you can install on a server. This guarantees that it will always run the same, regardless of the environment it is running in. So, instead of creating some instructions – you create a container.

How it works ?

On a high level, you have a Dockerfile for your project, and you it describes how container should be started. Docker has a good documentation, and there’s no sense on copying words from there. So, just refer to this part of documentation, and it should be clear for you.

How this Dockerfile looks like ?

How can it help me in my daily job ?

Imagin you have a new person on your project. You project is a complex system, which deploys not in a trivial way, and you have to show your new developer how he should start your project. Of course, you can say, that you can write a good README file, and your devs will be happy to read it and start working. And there’ll be no failures or errors. True story, what else to say.
altInstead of introducing your project in such insufficient way, delegate this work docker. Let him know how to start everything.

We love os x. How to work with docker on it ?

The recommended way to work with Docker is to use Docker Toolbox.

Okay, if I’ll fully rely on docker, how can I use it in all my services, including load balancers, etc ?

You can use docker-compose. The idea is that you write one file called docker-compose.yml and describe which services do you need. You can describe all the containers, that you need, including load balancers, service discoveries, etc…

Here’s how docker-compose.yml should look like:

web:  
  build: .
  ports:
   - "8000:8000"
  links:
   - redis
redis:  
  image: redis

Take a look at following line

build .

It means, that web is a custom container, and you should look for Dockerfile in the same folder, from which you start your compose.

Another interesting part of compose file is links. What are they ?

then you do

docker-compose up

under root folder, which contains docker-compose.yml and it should
build your system.

PS – Following article works fine as an example.

alt

Docker Compose. What beast is it ?

Can I run docker on production servers ? How? Is it worth doing that ?

I’ll explore all the problems of docker on production in following article.

 

Original blogpost

Seasoned software engineer. I've worked in many challenging projects. For now, I work as a part of Upwork engineering team. As a software engineer I do things in Java. Using Intelij products. A true fan of CI/CD. Truly believe, that nowadays machines cost less, than developers, that's why I'm always willing to automate everything. Because It works. And saves time. And money. Opensource commiter.