2

I have a lot of git repositories on my linux (ubuntu) server. I access all of them with the Apache user www-data. This worked great in the past because I was the only web developer. Now I have more developers and I need to restrict their access to some repositories. How do I modify my server/permissions/git repositories to accomodate this?

Let me show you the current process I go through to set up a repository.

To set up a new repository for the project project on myserver.com, I issue the following commands from terminal:

ssh [email protected]
password: (password for www-data)
cd /var/lib/repositories/
mkdir project.git
cd project.git
git init --bare
git update-server-info
cd /var/www/git.myserver.com/
ln -s /var/lib/repositories/project.git project

To deploy the project to the url http://project.myserver.com, I issue the following commands from terminal:

ssh [email protected]
password: (password for project)
cd /var/www/project.myserver.com/public_html/
git init
git remote add origin http://[email protected]/project
git pull origin master
password: (password for www-data)

The user project needs to know the password for www-data to push and pull from the git repository.

How do I modify my server/permissions/gitrepositories so that I can push and pull as project instead of www-data?

Additional Notes

Here's what the vhost entry for git.myserver.com looks like

<VirtualHost 173.255.230.136:80>
     ServerAdmin [email protected]
     ServerName git.myserver.com
     DocumentRoot /var/www/git.myserver.com/public_html/
     ErrorLog /var/www/git.myserver.com/logs/error.log
     CustomLog /var/www/git.myserver.com/logs/access.log combined
</VirtualHost>

The directory /var/lib/git is owned by the www-data user. This way, apache can write to the repos via the symbolic link in /var/www/git.myserver.com/public_html/project via the url http://[email protected]/project

2
  • What's the web server's configuration look like for the git.myserver.com host? And why are you logging on as www-data? Jan 7, 2012 at 6:14
  • Hi Shane, I answered your question in the additional notes.
    – John
    Jan 7, 2012 at 15:16

1 Answer 1

0

First I would have created a separate user for the git login. However that aside, probably the easiest way to manage this is to use something like gitosis which is a simple setup that allows you to do authenticated git access with granular permissions.

This will allow you to give people:

No access, read only, read write. Etc

http://scie.nti.st/2007/11/14/hosting-git-repositories-the-easy-and-secure-way

N.B. just checked the developers site and there are more features in this. Worth looking at for sure https://github.com/sitaramc/gitolite/

2
  • I have gitosis installed. What I'm afraid about is I don't understand the git clone command. I've been making new directories and using git init --bare to create new repositories. What are the consequences when I use git clone? And how do I convert existing repositories into a state that appears to have been git-clone?
    – John
    Jan 7, 2012 at 15:22
  • When you use git clone you copy the repository from the remote. So you'd want to use clone on existing repositories. git init is for creating new repositories which you then have to add the remote to. If you have existing repositories you should use git clone as init will not be using the existing repo it will be generating a new configuration in your current working directory.
    – haswalt
    Jan 9, 2012 at 17:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .