CODEWITHSUNDEEP
ruby-on-railscapistrano
Dan Rubio
Dan Rubio

Reputation: 4907

Why does Capistrano remove .git repository upon deployment?

I am deploying a web app with the following specific task under lib/capistrano/setup.rake file:

 69   desc "Publish git commit log to changelog.txt"
 70   task :publish_changelog do
 71     on roles(:web) do
 72       log = "#{deploy_to}/current/public/changelog.txt"
 73       execute "echo '============ Last deployed on' `date` ============\n > #{log}"
 74       execute "echo >> #{log}"
 75       execute "cd #{deploy_to}/current && git log --since=\"10 days ago\" >> #{log}"
 76     end
 77   end

I understand that this is a specific task only to my app, but the problem that I am having is that when I try to run this command, it breaks because after I deploy my application, Capistrano removes the .git directory on my application located on the server role(:web).

The command above works only if the application under role (:web) is a git repository. Essentially, what this means is that I have to ssh into the web server, remove the application "current" that previously was a git repository and then proceed to clone it again. Once that is done, running the above Capistrano task works. Why is that so? I must be missing something.

Here is what I have deduced. When I deploy a Capistrano app, Capistrano must be replacing my "current" with the git directory and pasting over a new one effectively removing the .git direcoty. Although Capistrano creates a shared and releases directory, it has to be pasting over my current one. Is this what Capistrano does? A lead to go on would be appreciated. I'm trying to decide if the command needs to be rewritten or if I need to somehow configure Capistrano. Thanks.

Upvotes: 1

Views: 881

Answers (1)

user777337
user777337

Reputation:

Capistrano keeps the cloned repository in #{deploy_to}/repo. There's a nice path variable for that: repo_path.

Notice a git repository at that path is actually a "bare repo".

Having it this way makes sense performance-wise because repo is cloned to this path only once. After that, the repo is only updated which is faster, especially for large repos. Cloning a git repository to current_path on each deploy would take longer.

That said, here's the stripped down version of your task that should work:

desc "Publish git commit log to changelog.txt"
task :publish_changelog do
  on roles(:web) do
    within repo_path do
      execute :git, :log, "--since='1100 days ago'"
    end
  end
end

Upvotes: 3

Related Questions