Hacking MAAS

Coding style

MAAS follows the Launchpad Python Style Guide, except where it gets Launchpad specific, and where it talks about method naming. MAAS instead adopts PEP-8 naming in all cases, so method names should usually use the lowercase_with_underscores form.

Prerequisites

You can grab MAAS’s code manually from Launchpad but Bazaar makes it easy to fetch the last version of the code. First of all, install Bazaar:

$ sudo apt-get install bzr

Then go into the directory where you want the code to reside and run:

$ bzr branch lp:maas maas && cd maas

MAAS depends on Postgres 9.1, RabbitMQ, Apache 2, daemontools, pyinotify, and many other packages. To install everything that’s needed for running and developing MAAS, run:

$ make install-dependencies

Careful: this will apt-get install many packages on your system, via sudo. It may prompt you for your password.

This will install bind9. As a result you will have an extra daemon running. If you are a developer and don’t intend to run BIND locally, you can disable the daemon by inserting exit 1 at the top of /etc/default/bind9. The package still needs to be installed for tests though.

You may also need to install python-django-piston, but installing it seems to cause import errors for oauth when running the test suite.

All other development dependencies are pulled automatically from PyPI when buildout runs. See First time using buildout? and Running tests.

First time using buildout?

Buildout is used to develop MAAS. Buildout, if configured so, can cache downloaded files and built eggs. If you’ve not already done something similar, the following snippet will massively improve build times:

[buildout]
download-cache = /home/<your-user-name>/.buildout/cache
eggs-directory = /home/<your-user-name>/.buildout/eggs

Put this in ~/.buildout/default.cfg and create the cache directory:

$ mkdir /home/<your-user-name>/.buildout/cache

The eggs directory will be created automatically.

Running tests

To run the whole suite:

$ make test

To run tests at a lower level of granularity:

$ ./bin/test.maas test src/maasserver/tests/test_api.py
$ ./bin/test.maas test src/maasserver/tests/test_api.py:AnonymousEnlistmentAPITest

The test runner is nose, so you can pass in options like --with-coverage and --nocapture (short option: -s). The latter is essential when using pdb so that stdout is not adulterated.

Running JavaScript tests

The JavaScript tests are run using Selenium. Firefox is the default browser but any browser supported by Selenium can be used to run the tests. Note that you might need to download the appropriate driver and make it available in the path. You can then choose which browsers to use by setting the environment variable MAAS_TEST_BROWSERS to a comma-separated list of the names of the browsers to use. For instance, to run the tests with Firefox and Chrome:

$ export MAAS_TEST_BROWSERS="Firefox, Chrome"

Running JavaScript tests with browsers on other platforms

The JavaScript tests can be run using the SauceLabs’ OnDemand service. There is a free version of this that provides 45 minutes a month of testing. To get started, sign up and go to your account page, select the Account tab, and click View my API Key. Now save your credentials:

$ mkdir -p ~/.saucelabs/connect
$ chmod go-rwx ~/.saucelabs/connect
$ echo "${username} ${api_key}" > ~/.saucelabs/connect/credentials

(You need to substitute your SauceLabs username and API key above.)

Next, like when running Selenium tests locally, you need to specify the browsers to run on. At the time of writing there are four to choose from, all running on a Windows host:

$ export MAAS_REMOTE_TEST_BROWSERS="IE7, IE8, IE9, Chrome"

By default, when MAAS_REMOTE_TEST_BROWSERS is not specified, testing via OnDemand is not attempted.

Development MAAS server setup

Access to the database is configured in src/maas/development.py.

The Makefile or the test suite sets up a development database cluster inside your branch. It lives in the db directory, which gets created on demand. You’ll want to shut it down before deleting a branch; see below.

First, set up the project. This fetches all the required dependencies and sets up some useful commands in bin/:

$ make

Create the database cluster and initialize the development database:

$ make syncdb

Optionally, populate your database with the sample data:

$ make sampledata

By default, the snippet maas_proxy includes a definition for an http proxy running on port 8000 on the same host as the MAAS server. This means you can either install squid-deb-proxy:

$ sudo apt-get install squid-deb-proxy

or you can edit contrib/snippets_v2/generic to remove the proxy definition.

Set the iSCSI config to include the MAAS configs:

$ sudo tee -a /etc/tgt/targets.conf < contrib/tgt.conf

The http_proxy variable is only needed if you’re downloading through a proxy; “sudo” wouldn’t pass it on to the script without the assignment. Or if you don’t have it set but do want to download through a proxy, pass your proxy’s URL: “http_proxy=http://proxy.example.com/”

Run the development webserver and watch all the logs go by:

$ make run

Point your browser to http://localhost:5240/

If you’ve populated your instance with the sample data, you can login as a simple user using the test account (username: ‘test’, password: ‘test’) or the admin account (username: ‘admin’, password: ‘test’).

At this point you may also want to download PXE boot resources.

To shut down the database cluster and clean up all other generated files in your branch:

$ make distclean

Downloading PXE boot resources

To use PXE booting, each cluster controller needs to download several files relating to PXE booting. This process is automated, but it does not start by default.

First create a superuser and start all MAAS services:

$ bin/maas-region-admin createadmin
$ make run

Substitute your own email. The command will prompt for a choice of password.

Next, get the superuser’s API key on the account preferences page in the web UI, and use it to log into MAAS at the command-line:

$ bin/maas login dev http://localhost:5240

Start downloading PXE boot resources:

$  bin/maas dev node-groups import-boot-images

This sends jobs to each cluster controller, asking each to download the boot resources they require. This may download dozens or hundreds of megabytes, so it may take a while. To save bandwidth, set an HTTP proxy beforehand:

$ bin/maas dev maas set-config name=http_proxy value=http://...

Running the built-in TFTP server

You will need to run the built-in TFTP server on the real TFTP port (69) if you want to boot some real hardware. By default, it’s set to start up on port 5244 for testing purposes. Make these changes:

* Edit ``etc/maas/pserv.yaml`` to change the tftp/port setting to 69
* Install the ``authbind``package:

  $ sudo apt-get install authbind

* Create a file ``/etc/authbind/byport/69`` that is *executable* by the
  user running MAAS.

  $ sudo touch /etc/authbind/byport/69
  $ sudo chmod a+x /etc/authbind/byport/69

Now when starting up the MAAS development webserver, “make run” and “make start” will detect authbind’s presence and use it automatically.

Running the BIND daemon for real

There’s a BIND daemon that is started up as part of the development service but it runs on port 5246 by default. If you want to make it run as a real DNS server on the box then edit services/dns/run and change the port declaration there so it says:

port=53

Then as for TFTP above, create an authbind authorisation:

$ sudo touch /etc/authbind/byport/53
$ sudo chmod a+x /etc/authbind/byport/53

and run as normal.

Running the cluster worker

The cluster worker process also needs authbind as it needs to bind a socket on UDP port 68 for DHCP probing:

$ sudo touch /etc/authbind/byport/68
$ sudo chmod a+x /etc/authbind/byport/68

If you omit this, nothing else will break, but you will get an error in the cluster celeryd log because it can’t bind to the port.

Configuring DHCP

MAAS requires a properly configured DHCP server so it can boot machines using PXE. MAAS can work with its own instance of the ISC DHCP server, if you install the maas-dhcp package:

$ sudo apt-get install maas-dhcp

If you choose to run your own ISC DHCP server, there is a bit more configuration to do. First, run this tool to generate a configuration that will work with MAAS:

$ maas-provision generate-dhcp-config [options]

Run maas-provision generate-dhcp-config -h to see the options. You will need to provide various IP details such as the range of IP addresses to assign to clients. You can use the generated output to configure your system’s ISC DHCP server, by inserting the configuration in the /etc/dhcp/dhcpd.conf file.

Also, edit /etc/default/isc-dhcp-server to set the INTERFACES variable to just the network interfaces that should be serviced by this DHCP server.

Now restart dhcpd:

$ sudo service isc-dhcp-server restart

None of this work is needed if you let MAAS run its own DHCP server by installing maas-dhcp.

Development services

The development environment uses daemontools to manage the various services that are required. These are all defined in subdirectories in services/.

There are familiar service-like commands:

$ make start
$ make status
$ make restart
$ make stop

The latter is a dependency of distclean so just running make distclean when you’ve finished with your branch is enough to stop everything.

Individual services can be manipulated too:

$ make services/pserv/@start

The @<action> pattern works for any of the services.

There’s an additional special action, run:

$ make run

This starts all services up and tails their log files. When you’re done, kill tail (e.g. Ctrl-c), and all the services will be stopped.

However, when used with individual services:

$ make services/webapp/@run

it does something even cooler. First it shuts down the service, then it restarts it in the foreground so you can see the logs in the console. More importantly, it allows you to use pdb, for example.

A note of caution: some of the services have slightly different behaviour when run in the foreground:

• Django (the webapp service) will be run with its auto-reloading enabled.
• Apache (the web service) will run with -X, which puts it in debug mode: only one worker will be started and the server will not detach from the console.

There’s a convenience target for hacking Django that starts everything up, but with Django in the foreground:

$ make run+webapp

Apparently Django needs a lot of debugging ;)

Adding new dependencies

Since MAAS is distributed mainly as an Ubuntu package, all runtime dependencies should be packaged, and we should develop with the packaged version if possible. All dependencies, from a package or not, need to be added to setup.py and buildout.cfg, and the version specified in versions.cfg (allowed-picked-version is disabled, hence buildout must be given precise version information).

If it is a development-only dependency (i.e. only needed for the test suite, or for developers’ convenience), simply running buildout like this will make the necessary updates to versions.cfg:

$ ./bin/buildout -v buildout:allow-picked-versions=true

Adding new source files

When creating a new source file, a Python module or test for example, always start with the appropriate template from the templates directory.

Database schema changes

MAAS uses South to manage changes to the database schema.

Be sure to have a look at South’s documentation before you make any change.

Changing the schema

Once you’ve made a model change (i.e. a change to a file in src/<application>/models/*.py) you have to run South’s schemamigration command to create a migration file that will be stored in src/<application>/migrations/.

Note that if you want to add a new model class you’ll need to import it in src/<application>/models/__init__.py

Once you’ve changed the code, run:

$ ./bin/maas-region-admin schemamigration maasserver --auto description_of_the_change

This will generate a migration module named src/maasserver/migrations/<auto_number>_description_of_the_change.py. Don’t forget to add that file to the project with:

$ bzr add src/maasserver/migrations/<auto_number>_description_of_the_change.py

To apply that migration, run:

$ make syncdb

Performing data migration

If you need to perform data migration, very much in the same way, you will need to run South’s datamigration command. For instance, if you want to perform changes to the maasserver application, run:

$ ./bin/maas-region-admin datamigration maasserver description_of_the_change

This will generate a migration module named src/maasserver/migrations/<auto_number>_description_of_the_change.py. You will need to edit that file and fill the forwards and backwards methods where data should be actually migrated. Again, don’t forget to add that file to the project:

$ bzr add src/maasserver/migrations/<auto_number>_description_of_the_change.py

Once the methods have been written, apply that migration with:

$ make syncdb

Documentation

Use reST with the convention for headings as used in the Python documentation.