Capistrano 3.1.x Strategy to deploy git projects with submodules.

submodule_strategy.rb

# Usage: 
# 1. Drop this file into lib/capistrano/submodule_strategy.rb
# 2. Add the following to your Capfile:
#   require 'capistrano/git'
#   require './lib/capistrano/submodule_strategy'
# 3. Add the following to your config/deploy.rb
#   set :git_strategy, SubmoduleStrategy

module SubmoduleStrategy
  # do all the things a normal capistrano git session would do
  include Capistrano::Git::DefaultStrategy
  
  # check for a .git directory
  def test
    test! " [ -d #{repo_path}/.git ] "
  end

  # same as in Capistrano::Git::DefaultStrategy
  def check
    test! :git, :'ls-remote', repo_url
  end

  def clone
    git :clone, '-b', fetch(:branch), '--recursive', repo_url, repo_path
  end

  # same as in Capistrano::Git::DefaultStrategy
  def update
    git :remote, :update
  end

  # put the working tree in a release-branch,
  # make sure the submodules are up-to-date
  # and copy everything to the release path
  def release
    release_branch = fetch(:release_branch, File.basename(release_path))
    git :checkout, '-B', release_branch, 
      fetch(:remote_branch, "origin/#{fetch(:branch)}")
    git :submodule, :update, '--init'
    context.execute "rsync -ar --exclude=.git\* #{repo_path}/ #{release_path}"
  end
end

Thanks mate worked like a charm

I followed your instructions, but the include Capistrano::Git::DefaultStrategy line is causing an error.

/lib/capistrano/submodule_strategy.rb:11:in <module:SubmoduleStrategy>: uninitialized constant Capistrano::Git (NameError)

Any idea why?

I have the same error as t-anjan.

It requires capistrano 3.1.0 to work.

For line 37 I get Unknown switch -B

DEBUG [77d89e1b] Command: cd ~/PATH/repo && ( GIT_ASKPASS=/bin/echo GIT_SSH=/FULLPATH/git-ssh.sh /usr/bin/env git checkout -B 20140311200837 origin/master )

This code is run twice, which when I switch the flag to a -b it freaks out that the branch exists.

I've got

lib/capistrano/submodule_strategy.rb:24:in `clone'

DEBUG [d000b805]    fatal: destination path '/var/www/application/repo' already exists and is not an empty directory.

@sld
from your server command line
sudo rm -rf /var/www/application/repo
Then from your workstation, deploy again via Capistrano

Your error occurs when using this gist on top of an existing project. This gist uses a different git strategy than ootb Capistrano 3.

Great work. I forked your fork to easily exclude files from deployment (such as lib, config, etc.).

Check it out here: https://gist.github.com/cickes/10388315

You have to escape the rsync command to make this work on ZSH

context.execute %|rsync -ar --exclude=".git\*" #{repo_path}/ #{release_path}|

Can't seem to find a way to make this work.

Tasks: TOP => git:create_release => git:update => git:clone

** Invoke deploy:failed (first_time) ** Execute deploy:failed

Deploying any other branch breaks with this message:

fatal: Needed a single revision

We can fix by adding this method:

def fetch_revision
  context.capture(:git, "rev-parse --short HEAD")
end

The default strategy does context.capture(:git, "rev-parse --short fetch(:branch)") but this breaks since we're making a new branch with the release timestamp.

Requiring 'capistrano/git' in your Capfile means that the git.rake is loaded twice, which leads to every Rake task in git.rake being defined twice, which leads to all of the git commands being run twice. Using the -B flag to force the second checkout to overwrite the first is a nasty hack to avoid the errors this throws. It would be better if this could be used without the extra require.

I realise that this is probably Capistrano's issue.

I'm going to try this solution unless there are any others recently..?

https://gist.github.com/seenmyfate/11206162