docs/manual/source/install/install-vagrant.html.md.erb

---
title: Installing PredictionIO with Vagrant (VirtualBox)
---

WARNING: Running PredictionIO with Vagrant is intended for the purposes of
simple tests in an isolated environment. Due to resource limitation and overhead
of virtual machine (VM), it runs much more slowly or may encounter memory issue.
We recommend using Linux or Mac machine for serious usage.

## Install VirtualBox

If you don't have VirtualBox installed, please follow the instructions in the
[VirtualBox site](https://www.virtualbox.org/wiki/Downloads) to download and
install it.

## Install Vagrant

If you don't have Vagrant installed, please follow the instructions in the the
[Vagrant site](https://www.vagrantup.com/downloads.html) to download and install
it.

## Bring up PredictoinIO VM with Vagrant

Get the latest vagrant setup from github and make sure in master branch:

```
$ git clone https://github.com/PredictionIO/PredictionIO-Vagrant.git
$ cd PredictionIO-Vagrant/
$ git checkout master
```

You may modify the following line in the Vagrantfile to change the memory and
CPU configuration setting of the VM, default is 2048 MB memory:

```
  v.customize ["modifyvm", :id, "--cpuexecutioncap", "90", "--memory", "2048"]
```

In the directory PredictionIO-Vagrant/, bring up PredictionIO VM by running:

```
$ vagrant up
```

INFO: When you run 'vagrant up' for the first time, it will download the base
box ubuntu/trusty64 if you don't have it. Then it will also install all
necessary libraries and setup PredictionIO in the virtual machine.

When it finishes successfully, you should see somthing like the following:

```
==> default: Installation done!
==> default: --------------------------------------------------------------------------------
==> default: Installation of PredictionIO <%= data.versions.pio %> complete!
==> default: IMPORTANT: You still have to start PredictionIO and dependencies manually:
==> default: Run: 'pio-start-all'
==> default: Check the status with: 'pio status'
==> default: Use: 'pio [train|deploy|...]' commands
==> default: Please report any problems to: support@prediction.io
==> default: Documentation at: http://docs.prediction.io
==> default: --------------------------------------------------------------------------------
==> default: Finish PredictionIO installation.
```

Now you have a PredictionIO VM running!

## Using the PredictionIO VM

### Login to the VM

You could ssh to the VM by running the following from your host machine in the
same directory where you run `vagrant up` (i.e. PredictionIO-Vagrant/)

```
$ vagrant ssh
```

Then your console prompt becomes something like:

```
vagrant@vagrant-ubuntu-trusty-64:~$
```

One you've logged into the VM, you can follow the one of the template quickstart
guides to get started.

To run PredictionIO Event Server and dependent services:

```
vagrant@vagrant-ubuntu-trusty-64:~$ pio-start-all
```

To deploy engine:

```
vagrant@vagrant-ubuntu-trusty-64:~/MyRecommendation$ pio deploy
```

### Shutdown and bring up PredictionIO VM again

You should shut down VM properly when you are not using PredictionIO:

```
$ vagrant halt
```

WARNING: If you didn't shut down VM properly or you ran `vagrant suspend`, the
VM may go to suspend state. HBase may not be running propoerly next time when
you run `vagrant up.` In this case, you can always run `vagrant halt` to do a
clean shutdown first before run `vagrant up` again.

Then you can run `vagrant up` again later to bring up the PredicitonIO VM again.

```
$ vagrant up
```

When it's ready, you should see the following:

```
==> default: --------------------------------------------------------------------------------
==> default: PredictionIO VM is up!
==> default: You could run 'pio status' inside VM ('vagrant ssh' to VM first) to confirm if PredictionIO is ready.
==> default: IMPORTANT: You still have to start the eventserver manually (inside VM):
==> default: Run: 'pio eventserver'
==> default: --------------------------------------------------------------------------------
```