Capistrano-offroad

Capistrano-offroad is a support package for using Capistrano with non-rails projects. It contains basic reset of Rails-specific tasks, a handful of utility functions, and modules with recipes.

Capistrano-offroad is available on terms of BSD license. See LICENSE file for details.

1 Installation

1.1 System-wide installation

Capistrano-offroad is available as a Ruby Gem, so in most cases you should simply call:

gem install capistrano-offroad

This requires fairly recent version of Ruby gems (1.3.6 or later), so it’s possible you’ll need to upgrade rubygems itself:

gem update --system

You can also download source code directly from Github page http://github.com/mpasternacki/capistrano-offroad and install it by rebuilding the gem manually.

1.2 Project-local installation

You can download source code from Github and use it as a Git submodule, or just drop it into your code base in whatever way suits you. Then, you should add to you Capfile the magic line adding capistrano-offroad’s lib/ subdirectory to the require path.

$:.unshift File.join( File.dirname(__FILE__), 'relative/path/to/capistrano-offroad/lib' )

2 Usage

When you’ve got capistrano-offroad on your require path, whether globally from a gem, or by extending require path, you simply require it.

require 'capistrano-offroad'

For some reason, this line should appear after the set :application declaration.

After that, you either require desired modules yourself, or use offroad_modules helper function.

require 'capistrano-offroad/modules/defaults'
require 'capistrano-offroad/modules/django'

offroad_modules 'defaults', 'django'

3 Features

3.1 Reset

By default, capistrano-offroad empties Rails-specific tasks: deploy:migrate, deploy:start, deploy:stop, deploy:restart.

The :deploy:finalize_update task is cut down to only run chgrp and chmod when :group_writable setting is true. chgrp behaviour can be modified by setting :group_writable to:

• :no_chgrp - don’t run chgrp command
• :sudo_chgrp - use sudo to make sure chgrp has proper permissions

The setting :shared_children is set to an empty list. This gives comfortable enough blank slate for working on non-rails projects, without redefining all the workflow that is already present in Capistrano.

3.2 Utilities

By default, two extensions to Capistrano are defined.

3.2.1 :run dependency type

Usage:

depend :remote, :run, "full shell command"=

When running cap deploy:check, run the supplied shell command. If its exit status is non-zero, the dependency has failed.

3.2.2 set_from_env_or_ask

Usage:

set_from_env_or_ask :setting_name, "Query prompt: "

If there is SETTING_NAME (uppercased first argument) in environment (i.e. supplied from command line), set :setting_name to value of this environment variable.

If the environment variable is not defined, ask user interactively for a value, using query prompt supplied as a second argument.

3.2.3 offroad_modules

Loads capistrano-offroad modules named as arguments. Convenience function, so that user doesn’t have to type require lines by himself.

3.3 Version

Three constants defining version of capistrano-offroad are defined:

CapistranoOffroad::VERSION::MAJOR
CapistranoOffroad::VERSION::MINOR
CapistranoOffroad::VERSION::TINY

Convenience string constant with full version number, separated with dots, is provided:

CapistranoOffroad::VERSION::STRING

3.3.1 require_version

A helper function, which can be used in Capfile to enforce a minimum version of capistrano-offroad:

CapistranoOffroad::VERSION.require_version(major, minor=0, tiny=0)

3.4 Modules

Modules are packs of recipes and helpers for dealing with specific software.

To load a module, either use the require function:

require 'capistrano-offroad/modules/module-name'

or use supplied offroad_modules helper function:

offroad_modules "foo", "bar", "baz", ...

When using a module, it is advised to at least browse through its source to know what to expect.

Following modules are defined:

3.4.1 defaults

Contains my personal defaults for settings. For me it’s part of reset, but I moved it out to a separate module, because other users may have different opinions. The settings are:

set :scm, :git
set :ssh_options, { :forward_agent => true }
set :use_sudo, false
set :deploy_to, "/srv/#{application}"

3.4.2 django

Settings, utilities and recipes for deploying Django projects.

3.4.2.1 Settings

Following settings are defined.

3.4.2.1.1 :python

Python interpreter command. Defaults to “python”.

Module also adds a dependency on this command to deploy:check.

3.4.2.1.2 :django_project_subdirectory

Directory in your repository that contains Django project (one with the the manage.py file). Defaults to “project”.

3.4.2.1.3 :django_use_south

Assume South is used for database migrations. Defaults to false.

3.4.2.1.4 :django_databases

This setting can be set to a list of database names (for Django-1.2 multi-database support), against which Capistrano will run syncdb/migrations. If unset or false, syncdb/migrations will be run with default database only (as for pre-1.2 Django).

3.4.2.2 Functions and utilities

django_manage cmd, options={}

Runs the manage.py command cmd. Optional keyword argument :path provides path to a Capistrano release where the command should be run. It defaults to :latest_release.

:python_module dependency type is defined to name Python modules in deploy:check dependencies:

depend :remote, :python_module, "MySQLdb"

3.4.2.3 Tasks

3.4.2.3.1 django:manage

Run custom Django management command in latest release.

Pass the management command and arguments in COMMAND=”…” variable. If COMMAND variable is not provided, Capistrano will ask for a command.

3.4.2.3.2 deploy:migrate

Runs manage.py syncdb in the latest release. If :django_use_south is true, it --migrate switch is used.

3.4.2.3.2.1 TODO make it run in specified release, as vanilla Rails deploy:migrate

3.4.2.3.3 TODO separate python module

Stuff that is not Django-specific should be moved to a separate python module that could be used with Python-related non-Django projects. The python module would be automatically loaded by django module.

3.4.3 supervisord

Control processes with the supervisor daemon.

3.4.3.1 Settings

set :supervisord_path, ""     # directory where supervisord binaries reside
set :supervisord_command, "supervisord"
set :supervisorctl_command, "supervisorctl"
set :supervisord_conf, "supervisord_conf"
set :supervisord_pidfile, "supervisord.pid"
set :supervisord_start_group, nil # process group to start on deploy:start - nil means all processes
set :supervisord_stop_group, nil  # process group to stop on deploy:stop - nil means all processes

3.4.3.2 Utilities

supervisorctl(cmd, options={})

Run a supervisorctl command specified as first argument. If optional keyword argument :try_start is true (the default, you may specify as false), start supervisord if not already running.

3.4.3.3 Tasks

3.4.3.3.1 Standard tasks

Stock Capistrano tasks deploy:start, deploy:stop, deploy:restart are defined. They optionally accept GROUP or PROGRAM environment variables, which can specify a single, specific process group or a single process name to start, stop or restart.

3.4.3.3.2 deploy:status

Runs supervisorctl status.

3.4.3.3.3 deploy:processes

Runs pstree with root in supervisord if supervisord is running.

3.4.3.3.4 deploy:reload_supervisord

Reloads supervisor daemon’s config. Starts it if not started.

3.4.3.3.5 deploy:run_supervisorctl

Runs supplied supervisorctl command. Command should be provided in COMMAND variable; if no variable is provided, capistrano-offroad will ask for the command.

3.4.4 daemontools

Recipes and settings for controlling processes with Dan Bernstein’s daemontools. It expects that run script lives in top level of your repository.

WARNING: this is legacy code, used only in old projects and not very well supported.

Also, this is probably not the best way to deal with the problem - it would be easier to manage, start and restart with the run script in the :deploy_to root, which would cd to current and exec ./run. If anybody wants to write a patch to support that, it would be great :)

3.4.4.1 Settings

set :svscan_root, "/service"
set :supervise_name, "#{application}"

3.4.4.2 Utilities

svc function runs svc with supplied arguments for the supervised directory.

3.4.4.3 Tasks

• daemontools:create_symlink - creates a symlink in :svscan_root
• daemontools:remove_symlink - removes symlink from :svscan_root and stops the app
• daemontools:status - displays svstat output for current release
• daemontools:relstatus - displays svstat output for all releases (use it after restart to make sure that previous release has actually stopped)
• deploy:start, deploy:stop, deploy:restart - standard Capistrano tasks
• deploy:symlink internal target is redefined; on_rollback handler is currently broken

3.4.5 monit

Module to run processes with monit. This is legacy, unsupported, unused code. Use it (and monit) on your own risk. I’m leaving it in only because the code is already written and it would be waste to throw it away only because monit sucks.

3.4.5.1 Settings

set :monit_group, nil         # process group to start/stop/reload
set :monit_command, "monit"

3.4.5.2 Tasks

• deploy:start, deploy:stop, deploy:restart - standard Capistrano tasks. They accept an extra optional variable PROCESS which, if given, specifies a single process to start, stop or restart. If PROCESS is not given, all processes (of :monit_group group, if configured) are started/stopped/restarted.
• deploy:status - displays monit summary output
• deploy:status:full - displays monit status output
• deploy:reload_monit - reloads monit configuration