I have gotten a few private queries about backups lately, and thought it would be good to say a few words about what and how we keep your repositories safe.
Patch-tag is hosted on linode, and we do backups using S3 in the amazon cloud.
Patch-tag backups up all repository and application state data once a day using backup manager. Archive files, which are in dar format, are split in 25 megabyte chunks before uploading to S3. Backups are full, not incremental, so every day is a complete snapshot of all repositories along with application (happpstack) state.The policy is to keep at least one month of backups.
After backups complete, I get a daily email stating the size of the data loaded into S3, along the lines of
20100312: backups look ok
application state: 0.80Gb
In case of errors creating archive files or uploading, I get a scary email saying that something went wrong. (Haven’t gotten one yet though.)
As a sanity check, I have done two full recreations of patch-tag from backups, using the EC2 amis I described in an earlier blog post. [ link please] I am happy to report there were no unpleasant surprises.
As of now, backups are sitewide. So you can’t access your personal deleted repos, but in case of emergency you could make a request for this data, assuming it is less than a month ago. We do have plans to change this in the near future to provide you more control over the backups of you repos.
If there are additional backup features you would like to see, please let me know.