Putting Git in the cloud with Amazon S3
I've been playing with the Git this week.
Trying it locally is all very well, but I now want to try it from various locations and computers. I guess I need a central repository...
Before signing up to GitHub or Unfuddle, I thought it might be fun to see if I could use it with Amazon's S3 cloud storage.
I only just got this set up, but it appears to be working!
In fact, I cheated, and used JungleDisk which gives me a file system wrapper on to S3.
Basic steps as follows:
- Sign up for Jungle Disk
- Download client software and install
- JungleDisk gives you drive that links to S3 (my disk in the cloud!)
- Now we just use standard git commands to set up the remote repository (thanks to Toolman Tim for nice instructions)
- Use
git --bear init ...to set up a blank repository on the cloud disk - Use
git remote add ...thengid push ...to move local repos to cloud disk. - On some other machine, also install JungleDisk. Then do a
git clone ...to get a working copy of the code.
I like this idea because
- Amazon S3 is cheap
- Amazon is low maintenance and has fail-over (unlike my Linux server)
- JungleDisk works on Mac, Linux and Windows
- I own the S3 account (no JungleDisk)
- I can still access the S3 using other tools and APIs (such as AWS::S3)
Disclaimer!
I don't know enough about Git or S3 to know if this is a good idea! Anyone got any thoughts?
Comments
-
Wow I was just searching through and thinking about making the move from svn to git when I found your post. I already have jungledisk and an S3 account so this would be perfect for me. Thanks for the idea. Also nice theme, I use the same one on my personal site :).
-
Are you still doing this?
I’ve been doing SVN on JungleDisk/S3 for a little while now but have had huge problems with the SVN revisions being corrupted every few days. I’m not experienced with git but considering making the switch because of this – is it going well?
-
@Tzillz
Yup, I’m still doing this, it works pretty well.
On 4 occassions GIT pull or push failed. But there was no corruption. I just had to repeat the command a few times before it reported success. I think this was due to the local cache of JungleDisk catching up with the data on the cloud.
In all honesty, I think this approach is certainly workable, and very cheap! If I could get SSH access directly to S3 that would be better, as I could use GIT over SSH to push and pull directly without needing JungleDisk as a mediator.
Eventually I may chose to spend more money on private GIT hosting, which would probably make for a smoother experience and give me extra features (at a cost, of course).
-
This is interesting, how would it work with multiple collaborators? I would like to remove the JungleDisk requirement, too. Seems like a lot could be achieved with a handful of rake tasks.