[clug] database backups with git

[Robert Edwards]

On 08/06/12 12:15, Carlo Hamalainen wrote:
> On 08/06/12 11:46, Robert Edwards wrote:
>> Anyone else use a VCS for maintaining snapshots of backup files/logs?
>>
>> What is the downside?
>
> Most incremental backup tools have a way of deleting old increments
> (e.g. rdiff-backup has "--remove-older-than"). How would you do this if
> the target is git, where (as far as I know) the idea is to retain all
> history forever?
>
> I've always thought of VCS as a way of storing mainly text written by
> humans, but then again there are projects like git-annex [1] so maybe
> things are changing.
>

Good points, Carlo - I hadn't seen git-annex before, I'll look at that
more closely - it might solve one or two other problems I have.

As for deleting old increments, I now have a variation to create a new
git repository each month. I still get a big saving in space. Deleting
old "snapshots" of the database (or log files, or whatever) is not
really an option (at least not for 7 years...), but I can see that the
git repository will continue to grow (albeit at about 1/3 the rate of
the raw dump files, according to my metrics so far). Creating a new
repository each month or so makes archiving/removing (really) old data
easier/possible.

Cheers,

Bob Edwards.