Architecture
DevTools uses a containerization based approach to providing a project environment. Project services running as containers are combined with configuration for persistent data storage, DNS services and network manipulation which allows for easy access via names like service.project.vm rather than IP addresses.
The DevTools CLI facilitates setting up a VM to act as a container host on systems which can not host containers natively. It also configures the VM with the following:
- Filesystem sharing for high performance file access of the host machine
- Container initiation for DNS services
- Network and name resolution configuration
Layered with this core are a set of docker images which provide common project services such as a web server along with the ability to configure these services with commonly desired development options. Any docker image can be used if desired.
Docker Compose is used to control containers to provide a complete environment and a set of Yeoman based generators are used to help initial project setup to ensure all pieces fit together using the current best known options..
Containerization
In order to provide lightweight environments and tooling we use a technical approach called containers. Containers attempt to move the unit of environment from server to application. This allows separation of concerns between how an application is configured, how the containers communicate with each other, and where the containers are deployed. Take an advanced Drupal stack that includes Apache/PHP, MySQL, Memcache, and SOLR. With each component configured to run in its own container, the containers can all run on a single VM for local development and be spread across multiple servers for an optimized production deployment. Let us briefly touch on the technology we will be using and how it conceptually fits together.
Docker
The container implementation we use is called Docker which is explained in the intro: What is Docker? This is the way we capture environment units for our application/services and share them with everyone on the team. Environments are captured as images, similar to a VM, so when anyone runs that image they all start with the exact same set of files. For example, nearly every project needs a web server so we have a container image that can be run to provide that service. Our Docker images are also set up to allow for configuration adjustment to enable common development options.
devtools-cli
This is a project that glues all of the hosting aspects of these tools together into an easy to use unit. You can find devtools-cli in our BitBucket repository here. There are 2 basic services that devtools-cli provides:
Manage virtual machines for running containers
The devtools binary will manage the creation/configuration/upgrade/start/stop of boot2docker virtual machines (a.k.a Docker Hosts) via docker-machine. It ensures that the docker-machine virtual machine is the right version, is named correctly and configured to run efficiently within Virtualbox, VMWare Fusion or xhyve.
Nice DNS names and routing for running containers
Once there is a safe environment to run our containers we need a way to route traffic to them and provide easy to use/remember domain names to make accessing these services simple. Domain names for containers are set in the docker-compose yaml files using configuration that the dnsdock container reads to create a mapping between domain name and container.
We use dnsdock running as a container within the Docker Host. The dnsdock service, which listens on 172.17.0.1:53535. The dnsdock container resolves the *.vm domain names to the IP addresses of the containers.
Internal container names will look like web.openatrium.vm. All DevTools containers will carry the .vm extension for name resolution. There is additional information in DNS Resolution.
Docker Hub
Docker Hub is where container images are stored and retrieved when your local machine does not already have a copy of the requested container image. Docker Hub can be thought of like GitHub or BitBucket and Docker Hub images can be thought of as git repositories. We can make new versions of the images and they can be pushed and pulled to the Docker Hub service.
Container Configuration
If a container wants to offer configurable options it will document how to control it within the README or via Environment
variables in the Dockerfile itself. See our Apache / PHP Dockerfile
for an example. In this Docker Image, passing environment variables can override the PHP memory limit. Those variables
can either be passed on the command line when executing a docker run
command directly, or in the environment
section
of a docker-compose file. Documentation entries within the Common Tasks section provide additional approaches to
container configuration.