This article is part of the Continuous Integration, Delivery and Deployment series.
The previous article described several ways to implement Continuous Deployment. Specifically, it described, among other things, how to implement it using Docker to deploy applications as containers and nginx for reverse proxy necessary for successful utilization of blue-green deployment technique. All that was running on top of CoreOS, operating system specifically designed for running Docker containers.
Source code used in this article can be found in the GitHub repo vfarcic/provisioning (directory ansible).
Ansible is an open-source software platform for configuring and managing computers. It combines multi-node software deployment, ad hoc task execution, and configuration management. It manages nodes over SSH. Modules work over JSON and standard output and can be written in any programming language. The system uses YAML to express reusable descriptions of systems.
Preferable way to work with Ansible is with roles. They describe a set of tasks that should be run in order to setup something. In our case, we'll have 5 roles described in bdd.yml.
|- hosts: all|
First four roles (etcd, confd, docker and nginx) will make sure that the tools we need for blue-green deployment are present. Docker role, for example, is defined as following:
|- name: Docker is present|
|apt: name=docker.io state=present|
|- name: Python-pip is present|
|apt: name=python-pip state=present|
|- name: Docker-py is present|
|pip: name=docker-py version=0.4.0 state=present|
It installs Docker using apt-get, Python PIP and Docker-py. We need Docker to manage our containers. The Docker-py is a Python library required for the Ansible Docker module that we'll use to run nginx container.
As you can see, Ansible is very easy to use and understand. Just by reading YML files one can easily see what is going on. Simplicity is one of the main advantages it has over similar tools like Puppet and Chef. After very short introduction to Ansible, all we have to do is look for a module that performs tasks we need (i.e. apt for installation of Debian packages) and describe it as a YML task.
Once we have the tools installed, it's time to take a look at the last Ansible role bdd. This is where deployment happens. However, before we proceed, let me explain goals I had in mind before I started working on it.
Deployment should follow blue-green technique. While the application is running, we would deploy a new version in parallel with the old one. Container that will be used already passed all sorts of unit and functional tests giving us reasonable confidence that each release is working correctly. However, we still need to test it after deployment in order to make the final verification that what we deployed is working correctly. Once all post-deployment tests passed we are ready to make the new release available to the public. We can do that by changing our nginx reverse proxy to redirect all requests to the newly deployed release. In other words, we should do following.
- Pull the latest version of the application container
- Run the latest application version without stopping the existing one
- Run post-deployment tests
- Notify etcd about the new release (port, name, etc)
- Change nginx configuration to point to the new release
- Stop the old release
If we do all of the above, we should accomplish zero-downtime. At any given moment our application should be available.
On top of the procedure described above, deployment should work both with and without Ansible. While using it helps a lot, all essential elements should be located on the server itself. That means that scripts and configurations should be on the machine we're deploying to and not somewhere on a central server.
The role bdd is following.
|- name: TOML is present|
|- name: Config template is present|
|- name: Deployment script is present|
|- name: Deployment is run|
The first task makes sure that the template resource bdd.toml is present. It is used by confd to specify what is the template, what is the destination and is the command that should be executed (in our case restart of the nginx container).
The second task makes sure that the confd template (bdd.conf.tmpl)[https://github.com/vfarcic/provisioning/blob/master/ansible/roles/bdd/files/bdd.conf.tmpl] is present. This template together with the bdd.toml will change nginx proxy to point to a new release every time we deploy it. That way we'll have no interruption to our service.
The third task makes sure that the deployment script deploy-bdd.sh is present and the last one that it is run.
From the Ansible point of view, that's all there is. The real "magic" is in the deployment script itself. Let's go through it.
We start by discovering whether we should do blue or green deployment. If the current one is blue we'll deploy green and the other way around. Information about currently deployed "color" is stored in etcd key /bdd/color.
Once the decision is made, we stop and remove existing containers if there are any. Keep in mind that the current release will continue operating and won't be affected until the very end.
In this particular case a set of BDD scenarios are run using PhantomJS headless browser.
If all tests passed, we should store information about the new release using etcd and run confd that will update our nginx configuration. Until this moment, nginx was redirecting all the requests to the old release. If nothing failed in the process so far, only from this point will users of the application be redirected to the version we just deployed.
Finally, since we have the new release deployed, tested and made available to the general public through nginx reverse proxy, we're ready to stop and remove the old version.
Running it all together
Let's see it in action. I prepared a Vagrantfile that will create an Ubuntu virtual machine and run Ansible playbook that will install and configure everything and, finally, deploy the application. Assuming that Git, Vagrant and VirtualBox are installed, run following.
First run might take a while since Vagrant and Ansible will need to download a lot of stuff (OS, packages, containers...). Please be patient especially on slower bandwidth. Good news is that each consecutive run will be much faster.
To simulate deployment of a new version, run following.
If you SSH to the VM you can see that the running version changes from blue (port 9001) to green (port 9002) and the other way around each time we run vagrant provision.
Before, during and after deployment, the application will be available without any interruption (zero-downtime). You can check it out by opening http://localhost:8000/ in your favorite browser.
As you could see, deployment is greatly simplified with Docker and containers. While, in a more traditional setting, Ansible would need to install a bunch of stuff (JDK, web server, etc) and make sure that ever-increasing amount of configuration files are properly set, with containers the major role of Ansible is to make sure that OS is configured, that Docker is installed and that few other things are properly set. In other words, Ansible continues being useful while an important part of its work is greatly simplified with containers and the concept of immutable deployments (what we deploy is unchangeable). We're not updating our applications with new versions. Instead, we're deploying completely new containers and removing old ones.
All the code used in this article can be found in the directory ansible inside the GitHub repo vfarcic/provisioning.
Can we put all this into Jenkins server and apply Continuous Deployment process? That's the topic of the Continuous Integration, Delivery or Deployment with Jenkins, Docker and Ansible article.
The DevOps 2.0 Toolkit
If you liked this article, you might be interested in The DevOps 2.0 Toolkit: Automating the Continuous Deployment Pipeline with Containerized Microservices book.
This book is about different techniques that help us architect software in a better and more efficient way with microservices packed as immutable containers, tested and deployed continuously to servers that are automatically provisioned with configuration management tools. It's about fast, reliable and continuous deployments with zero-downtime and ability to roll-back. It's about scaling to any number of servers, design of self-healing systems capable of recuperation from both hardware and software failures and about centralized logging and monitoring of the cluster.
In other words, this book envelops the whole microservices development and deployment lifecycle using some of the latest and greatest practices and tools. We'll use Docker, Kubernetes, Ansible, Ubuntu, Docker Swarm and Docker Compose, Consul, etcd, Registrator, confd, Jenkins, and so on. We'll go through many practices and, even more, tools.