archive command is basically the equivalent of SVN’s
export – it dumps a copy of the entire repository without any of the version control files, making it perfect for deploying to a testing or production server.
Using a combination of
git archive, SSH, and Gzip; deploying a Git repository to a remote server can be done quickly and easily:
git archive --format=tar origin/master | gzip -9c | ssh firstname.lastname@example.org "tar --directory=/var/www -xvzf -"
Let’s take a look at the different parts of that command.
This tells Git to combine all of the repository’s files into a single tarball file. The benefit of doing this is that a large single file can be transferred much quicker than thousands of smaller files.
If you are familiar with Git, you should recognise that these are the default names for a Git remote and branch. You should change these to match your remote/branch setup.
The tarball created by
git archive is piped into Gzip, which applies compression to reduce the size of the file. The
9 flag tells Gzip to use the best compression method, and the
c flag makes Gzip write the compressed file to stdout. The reason we want Gzip to write to standard output is so that we can send the compressed repository straight to the remote server via SSH.
ssh email@example.com "tar --directory=/var/www -xvzf -"
The Gzipped tarball is then piped through SSH to your remote server. The remote server runs
tar --directory=/var/www -xvzf - which extracts the Gzipped tarball into the
/var/www directory. Note that the hyphen (
-) at the end of the
tar command tells tar to receive data from a piped command instead of a file.
…And that’s all there is to it! Using this command I am able to compress a 25MB repository down to 1MB and deploy it in less than 10 seconds. I hope you find it just as useful.