Capistrano

Requirements

• Ruby >= 1.9 (JRuby and C-Ruby/YARV are supported)

Support

Need help with getting Capistrano up and running? Got a code problem you want to get solved quickly?

Get Capistrano support on CodersClan.

Installation

Add this line to your application's Gemfile:

gem 'capistrano', '~> 3.2.0'

And then execute:

$ bundle install

Capify: make sure there's no "Capfile" or "capfile" present

$ bundle exec cap install

This creates the following files:

├── Capfile
├── config
│   ├── deploy
│   │   ├── production.rb
│   │   └── staging.rb
│   └── deploy.rb
└── lib
    └── capistrano
            └── tasks

To create different stages:

$ bundle exec cap install STAGES=local,sandbox,qa,production

Usage

$ bundle exec cap -T

$ bundle exec cap staging deploy
$ bundle exec cap production deploy

$ bundle exec cap production deploy --dry-run
$ bundle exec cap production deploy --prereqs
$ bundle exec cap production deploy --trace

Tasks

server 'example.com', roles: [:web, :app]
server 'example.org', roles: [:db, :workers]
desc "Report Uptimes"
task :uptime do
  on roles(:all) do |host|
    execute :any_command, "with args", :here, "and here"
    info "Host #{host} (#{host.roles.to_a.join(', ')}):\t#{capture(:uptime)}"
  end
end

Note:

tl;dr: execute(:bundle, :install) and execute('bundle install') don't behave identically!

execute() has a subtle behaviour. When calling within './directory' { execute(:bundle, :install) } for example, the first argument to execute() is a Stringish with no whitespace. This allows the command to pass through the SSHKit::CommandMap which enables a number of powerful features.

When the first argument to execute() contains whitespace, for example within './directory' { execute('bundle install') } (or when using a heredoc), neither Capistrano, nor SSHKit can reliably predict how it should be shell escaped, and thus cannot perform any context, or command mapping, that means that the within(){} (as well as with(), as(), etc) have no effect. There have been a few attempts to resolve this, but we don't consider it a bug although we acknowledge that it might be a little counter intuitive.

Before / After

Where calling on the same task name, executed in order of inclusion

# call an existing task
before :starting, :ensure_user

after :finishing, :notify


# or define in block
before :starting, :ensure_user do
  #
end

after :finishing, :notify do
  #
end

If it makes sense for your use case (often, that means generating a file) the Rake prerequisite mechanism can be used:

desc "Create Important File"
file 'important.txt' do |t|
  sh "touch #{t.name}"
end
desc "Upload Important File"
task :upload => 'important.txt' do |t|
  on roles(:all) do
    upload!(t.prerequisites.first, '/tmp')
  end
end

The final way to call out to other tasks is to simply invoke() them:

namespace :example do
  task :one do
    on roles(:all) { info "One" }
  end
  task :two do
    invoke "example:one"
    on roles(:all) { info "Two" }
  end
end

This method is widely used.

Getting User Input

desc "Ask about breakfast"
task :breakfast do
  ask(:breakfast, "pancakes")
  on roles(:all) do |h|
    execute "echo \"$(whoami) wants #{fetch(:breakfast)} for breakfast!\""
  end
end

Perfect, who needs telephones.

When using ask to get user input, you can pass echo: false to prevent the input from being displayed:

ask(:database_password, "default", echo: false)

Using password authentication

Password authentication can be done via set and ask in your deploy environment file (e.g.: config/deploy/production.rb)

set :password, ask('Server password', nil)
server 'server.domain.com', user: 'ssh_user_name', port: 22, password: fetch(:password), roles: %w{web app db}

Running local tasks

Local tasks can be run by replacing on with run_locally

desc 'Notify service of deployment'
task :notify do
  run_locally do
    with rails_env: :development do
      rake 'service:notify'
    end
  end
end

Of course, you can always just use standard ruby syntax to run things locally

desc 'Notify service of deployment'
task :notify do
  %x('RAILS_ENV=development bundle exec rake "service:notify"')
end

Alternatively you could use the rake syntax

desc "Notify service of deployment"
task :notify do
   sh 'RAILS_ENV=development bundle exec rake "service:notify"'
end

Console

Note: Here be dragons. The console is very immature, but it's much more cleanly architected than previous incarnations and it'll only get better from here on in.

Execute arbitrary remote commands, to use this simply add require 'capistrano/console' which will add the necessary tasks to your environment:

$ bundle exec cap staging console

Then, after setting up the server connections, this is how that might look:

$ bundle exec cap production console
capistrano console - enter command to execute on production
production> uptime
 INFO [94db8027] Running /usr/bin/env uptime on leehambley@example.com:22
DEBUG [94db8027] Command: /usr/bin/env uptime
DEBUG [94db8027]   17:11:17 up 50 days, 22:31,  1 user,  load average: 0.02, 0.02, 0.05
 INFO [94db8027] Finished in 0.435 seconds command successful.
production> who
 INFO [9ce34809] Running /usr/bin/env who on leehambley@example.com:22
DEBUG [9ce34809] Command: /usr/bin/env who
DEBUG [9ce34809]  leehambley pts/0        2013-06-13 17:11 (port-11262.pppoe.wtnet.de)
 INFO [9ce34809] Finished in 0.420 seconds command successful.

A word about PTYs

There is a configuration option which asks the backend driver to ask the remote host to assign the connection a pty. A pty is a pseudo-terminal, which in effect means tell the backend that this is an interactive session. This is normally a bad idea.

Most of the differences are best explained by this page from the author of rbenv.

When Capistrano makes a connection it is a non-login, non-interactive shell. This was not an accident!

It's often used as a band aid to cure issues related to RVM and rbenv not loading login and shell initialisation scripts. In these scenarios RVM and rbenv are the tools at fault, or at least they are being used incorrectly.

Whilst, especially in the case of language runtimes (Ruby, Node, Python and friends in particular) there is a temptation to run multiple versions in parallel on a single server and to switch between them using environmental variables, this is an anti-pattern, and symptomatic of bad design (e.g. you're testing a second version of Ruby in production because your company lacks the infrastructure to test this in a staging environment).

Configuration

The following variables are settable:

Variable Name	Description	Notes
:repo_url	The URL of your scm repository (git, hg, svn)	file://, https://, ssh://, or svn+ssh:// are all supported
:branch	The branch you wish to deploy	This only has meaning for git and hg repos, to specify the branch of an svn repo, set :repo_url to the branch location.
:scm	The source control system used	:git, :hg, :svn are currently supported
:tmp_dir	The (optional) temp directory that will be used (default: /tmp)	if you have a shared web host, this setting may need to be set (i.e. /home/user/tmp/capistrano).

Support removed for following variables:

Variable Name	Description	Notes
:copy_exclude	The (optional) array of files and/or folders excluded from deploy	Replaced by Git's native .gitattributes, see #515 for more info.

Host and Role Filtering

Capistrano enables the declaration of servers and roles, each of which may have properties associated with them. Tasks are then able to use these definitions in two distinct ways:

• To determine configurations: typically by using the roles(), release_roles() and primary() methods. Typically these are used outside the scope of the on() method.

• To interact with remote hosts using the on() method

An example of the two would be to create a /etc/krb5.conf' file containing the list of available KDC's by using the list of servers returned by roles(:kdc)and then uploading it to all client machines usingon(roles(:all)) do upload!(file) end`

A problem with this arises when filters are used. Filters are designed to limit the actual set of hosts that are used to a subset of those in the overall stage, but how should that apply in the above case?

If the filter applies to both the interaction and configuration aspects, any configuration files deployed will not be the same as those on the hosts excluded by the filters. This is almost certainly not what is wanted, the filters should apply only to the interactions ensuring that any configuration files deployed will be identical across the stage.

Another type of filtering is done by defining properties on servers and selecting on that basis. An example of that is the 'no_release' property and it's use in the release_roles() method. To distinguish these two types of filtering we name them:

• On-Filtering Specified in the following ways:
  • Via environment variables HOSTS and ROLES
  • Via command line options --hosts and --roles
  • Via the :filter variable set in a stage file
• Property-Filtering These are specified by options passed to the roles() method (and implicitly in methods like release_roles() and primary())

To increase the utility of On-Filters they can use regular expressions:

• If the host name in a filter doesn't match /^[-A-Za-z0-9.]+$/ (the set of valid characters for a DNS name) then it's assumed to be a regular expression.
• Since role names are Ruby symbols they can legitimately contain any characters. To allow multiple of them to be specified on one line we use the comma. To use a regexp for a role filter begin and end the string with '/'. These may not contain a comma.

When filters are specified using comma separated lists, the final filter is the union of all of the components. However when multiple filters are declared the result is the intersection.

SSHKit

SSHKit is the driver for SSH connections behind the scenes in Capistrano. Depending on how deep you dig, you might run into interfaces that come directly from SSHKit (the configuration is a good example).

Testing

Capistrano has two test suites: an RSpec suite and a Cucumber suite. The RSpec suite handles quick feedback unit specs. The Cucumber features are an integration suite that uses Vagrant to deploy to a real virtual server. In order to run the Cucumber suite you will need to install Vagrant and Vagrant supported virtualization software like VirtualBox.

# To run the RSpec suite
$ rake spec

# To run the Cucumber suite
$ rake features

# To run the Cucumber suite and leave the VM running (faster for subsequent runs)
$ rake features KEEP_RUNNING=1

License

MIT License (MIT)

Copyright (c) 2012-2013 Tom Clements, Lee Hambley

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.