This sets up a VM that works reasonably well with Stack's Docker integration if you're on a non-Linux platform. Seestack#194 for why boot2docker does not work with Stack's Docker integration. This is a work-in-progress and has rough edges. That said, it seems to work alright. The VM will also work for general Docker use without Stack.
Docker in the VM is configured to use the new
overlay storage driver. Despite its newness, I've had no problems with it, and every other driver has given me trouble when building large images, so I definitely recommend it.
This uses Vagrant's support for synced folders using NFS, which alleviates the extreme slowness of VirtualBox shared folders (see boot2docker/boot2docker#593), but is still significantly slower than native filesystem mounting.
Note: requires at least stack-0.1.6.0 and Docker 1.9.1
Note: only tested on Mac OS X. It definitely won't work on Windows, as it uses NFS, and Stack's Docker integration doesn't support Windows paths.
To set up:
- Edit the
Vagrantfileand adjust the constants at the top to your preference. - Run
vagrant up. You will probably have to enter your root password (on the host) so that Vagrant can set up its NFS exports. - Set the
DOCKER_HOSTenvironment variable:export DOCKER_HOST="tcp://192.168.83.84:2375"Adjust the IP address if you changed thePRIVATE_IP_ADDRESSconstant in theVagrantfile.
Now use you can use
stack with Docker enabled normally from your host. You can also use docker commands from the host.
To access server processes running in a Docker container (e.g.
stack exec warp), you must connect to the VM's IP address instead of localhost (e.g. open http://192.168.83.84:3000/).
from https://github.com/borsboom/vagrant-stack-docker
