Build your own PaaS

This document describes how to create a private PaaS service using tsuru. It contains instructions on how to build tsuru and some of its components from source.

This document assumes that tsuru is being installed on a Ubuntu machine. You can use equivalent packages for beanstalkd, git, MongoDB and other tsuru dependencies. Please make sure you satisfy minimal version requirements.

There’s also a contributed Vagrant box, that setups a PaaS using Chef. You can check this out:


The Tsuru PaaS is composed by multiple components:

  • tsuru server
  • tsuru collector
  • gandalf
  • charms

And these components have their own dependencies, like:

  • mongodb (>=2.2.0)
  • beanstalkd (>=1.4.6)
  • git-daemon (git>=1.7)
  • juju (python version, >=0.5)
  • libyaml (>=0.1.4)


1. Operating System

At the moment, tsuru server is fully supported and tested on Ubuntu 12.04 and the steps below will guide you throught the install process.

If you try to build tsuru server on most Linux systems, you should have few problems and if there are problems, we are able to help you. Just ask on #tsuru channel on

  • Have you tried tsuru server on other systems? Let us know and contribute to the project.

2. Hardware

Tsuru server is a lightweight framework and can be run in a single small machine along with all the deps.

3. Software

3.1 MongoDB

Tsuru needs MongoDB stable, distributed by 10gen. It’s pretty easy to get it running on Ubuntu

3.2 Juju

Tsuru uses juju to orchestrates your “apps”. To install juju follow the juju install guide. Please make sure that you configure Juju properly. Then run:

$ juju bootstrap

Juju Charms define how platforms will be installed. You may take a look at juju charms collection or use the charms provided by tsuru

Put it somewhere and define the setting juju:charms-path in the configuration file:

$ git clone git:// /home/me/charms
$ cat /etc/tsuru/tsuru.conf
# ...
  charms-path: /home/me/charms

3.3 Beanstalkd

Tsuru uses Beanstalkd as a work queue. Install the latest version, by doing this:

$ sudo apt-get install -y beanstalkd

3.4 Gandalf

Tsuru uses Gandalf to manage git repositories, to get it installed follow this steps

Installing pre-built binaries

You can download pre-built binaries of tsuru and collector. There are binaries available only for Linux 64 bits, so make sure that uname -m prints x86_64:

$ uname -m

Then download and install the tsr binary:

$ curl -sL | sudo tar -xz -C /usr/bin

These commands will install tsr in /usr/bin (you will need to be a sudoer and provide your password). You may install this command in your PATH.

Installing from source

  1. Build dependencies

To build tsuru from source you will need to install the following packages

$ sudo apt-get install -y golang-go git mercurial bzr gcc
  1. Install the tsuru tsr

Add the following lines to your ~/.bashrc:

$ export GOPATH=/home/ubuntu/.go
$ export PATH=${GOPATH}/bin:${PATH}

Then execute:

$ source ~/.bashrc
$ go get

Configuring tsuru

Before running tsuru, you must configure it. By default, tsuru will look for the configuration file in the /etc/tsuru/tsuru.conf path. You can check a sample configuration file and documentation for each tsuru setting in the “Configuring tsuru” page.

You can download the sample configuration file from Github:

$ [sudo] mkdir /etc/tsuru
$ [sudo] curl -sL -o /etc/tsuru/tsuru.conf

Make sure you define the required settings (database connection, authentication configuration, AWS credentials, etc.) before running tsuru.

Running tsuru

Now that you have tsr properly installed, and you configured tsuru, you’re three steps away from running it.

  1. Start mongodb
$ sudo service mongodb  start
  1. Start beanstalkd
$ sudo service beanstalkd start
  1. Start api and collector
$ tsr api &
$ tsr collector &

One can see the logs in:

$ tail -f /var/log/syslog

Using tsuru

Congratulations! At this point you should have a working tsuru server running on your machine, follow the tsuru client usage guide to start build your apps.