Nuxeo Cloud Controller

The project currently provides a "ncc" command-line tool that allows to create, start, stop and destroy Nuxeo instances on a single dedicated server (or VM).

Note: this command-line tool is supposed to be used on the server, so it's really useful for a sysadmin or to experiment with the system. It is not a command-line tools for developers (a la 'bees' or 'vmc').

What it does

• Manage instances (creates them from a "model", setup the config file, wire them behind nginx)
• Stop them after some inactivity (via a script launched from cron)
• When an instance is stopped, point nginx to a small webserver that will restart the instance when a web request is made to it

Install

You need Python and pip (maybe easy_install too).

Simply running "python setup.py install" should be enough, if you have installed all the dependencies (run "sudo pip install -r deps.txt" first).

You can also use "make env" to create a virtualenv, then activate it with "source env/bin/activate".

To test, you need to create a nuxeocloud directory in your home directory with a certain structure. You can use "sh setup.sh" for this.

You would a wildcard DNS for the system to work. Since that's not practical, you can add to your /etc/hosts file something like:

127.0.0.1 nuxeo1 nuxeo2 nuxeo3 nuxeo4 nuxeo5

(Assuming you will never test more than 5 instances).

Postgresql needs to be started.

You also need to be sudoer (on Linux). (Of course this will be relaxed in the future).

Using

The config is hardcoded in config.py. However, this is easy to fix.

Typical scenario is:

• ncc boot # starts nginx, mostly
• ncc create
• ncc stop iid
• ncc destroy iid

Currently, instance with iid (instance id) iid will be mapped to host "nuxeoiid". This can be changed in model.py.

Run "ncc help" for more info.

I parallel, you need to start the web server ("ncc server") and from time to time (i.e. using a crontab) to run "ncc monitor".

Architecture

We currently have:

• A small database (currently Sqlite,could be moved transparently to Postgres) that stores information about the instances
• A common Postgres database for the instances (currently needs to be started manually)
• A supervisord daemon that is responsible for starting / stopping the instances (and nginx)
• Nginx that acts as the proxy
• A small webserver to process requests to sleeping instances

TODO

Lots of things:

• Pass parameters to ncc create (like: admin email, hostname, etc.) and manage more finely
  • Instance names, database names, passwords, admin password
  • Port
• Better state control for instances
• One possibility is to move all the current scripts (only useful for testing) to the web server, an add a small command line tool that calls the web server via REST/JSON
• Make a nice admin interface using the web server
• Make an end-user interface using the web server ?
• Plug with Connect or Studio ?
• Create a command-line tool for the users (a la vmc) ? (Probably not, at least not for customers)

Cleanup:

• Ensure that everything is rock-solid.