Setup

The simplest way to work on SkillTree is via a virtual machine. These instructions should work on Linux or Mac - I'm not sure how on Windows, but should be possible.

First, install the following three packages, via whatever mechanism you have:

• virtualbox (tested on 4.3.12)
• vagrant (tested on 1.6.3)
• ansible (tested on 1.6.6)
• sshpass (for production provisioning only)

Next, move into the directory of the SkillTree source and run the following command:

vagrant up

It seems to sometimes fail out the first time after it starts to run the provisioner, so if you see that error, just run:

vagrant provision

Provisioning is idempotent, so you can do it over and over and nothing will change.

After that completes, you will have a running virtual machine with all the needed libraries and binaries. The only thing left to do is to configure the Makefile to use Vagrant. Copy .vagrantmode.example to .vagrantmode

Now run make init to configure the package, run make dbup to run all the database migrations, and run make run to build and start the application. It will start the application on port 8000, so once it starts up, visit localhost:8000 in a web browser to see the application, and when you edit the source files or templates, it'll automatically recompile.

Production

To set up a production deployment, first set the IP address of the server (should be debian 7.5) in both provisioning/inventory and the Makefile. You should already have the VM set up as above (we use the VM to build production binaries. Not using the VM is not supported if you want to deploy). First create the production config files. The production database password is in the ansible vault. So first get the password.txt file by some means, put it in the provisioning folder, and open up the vault with

ansible-vault edit --vault-password-file=provisioning/password.txt provisioning/secrets.yml

Take the database password and use it to create the prod.cfg file in snaplets/persist and snaplets/postgresql-simple, modeling off of the prod-sample.cfg files.

If this is a new production server, run make production-init. It'll prompt you for the root password on the server (a few times). If the server has already been set up, this step isn't needed.

Run database migrations with make dbup PRODUCTION=1, and deploy the application with make deploy. Whenever you want to deploy a new version, just run make deploy. If there is an error on startup, the new version will not be switched to, and you can debug it by looking at either the application logs (run make production-log) or the logs for the managing process (run make production-keter-log).

Haskell on your host machine

If you want to also have haskell on your development machine (in addition to the VM), or set up development (not done yet), you can also use ansible to do the same installs (right now only ghc and cabal). Run

ansible-playbook -i provisioning/inventory provisioning/native.yml

From the project directory (or from anywhere else, changing the paths to inventory and native.yml to be correct).

Running tests

When you set up the project, it'll download selenium and the chrome driver. Provided you have PROJDIR/.cabal-sandbox/bin in your PATH, make test should set up and run the selenium tests, including starting the server.