threads / discuss / 11036

Removing old data without disturbing tree?

Subject: Removing old data without disturbing tree?

## tl;dr

3 messages between Nov 27, 2007 and Nov 27, 2007.

replies: 2people: 2as markdown or json

David Brown· Nov 27, 2007, 19:39 UTC · lore

An upstream tree I'm mirroring with git-p4 has decided to start checking in large tarballs (150MB) periodically. It's basically a prebuild version of some firmware needed to run the rest of the software.

Git doesn't seem to have any problem with these tarballs (and is using a lot less space than P4), but I have a feeling we might start running into problems when things get real big. Does anyone have experience with packs growing beyong several GB?

Aside from that, is there an easy way to prune out the old history from a working tree? I'd like something like what 'git clone --depth n' would produce, and I suppose I could do the clone and then pivot the trees. I mainly don't want to be rewriting history, just making parts inaccessible.

Thanks, David Brown

Nicolas Pitre· Nov 27, 2007, 20:06 UTC · re: David Brown · lore

Re: Removing old data without disturbing tree?

On Tue, 27 Nov 2007, David Brown wrote:
Show 8 quoted lines
> An upstream tree I'm mirroring with git-p4 has decided to start checking
> in large tarballs (150MB) periodically.  It's basically a prebuild version
> of some firmware needed to run the rest of the software.
> 
> Git doesn't seem to have any problem with these tarballs (and is using a
> lot less space than P4), but I have a feeling we might start running into
> problems when things get real big.  Does anyone have experience with packs
> growing beyong several GB?

It should just work. It was tested with artificial data sets but that's about it.

Now if those tarballs are actually multiple revisions of the same package, you might consider storing them uncompressed and let Git delta compress them against each other which will produce an even more significant space saving.

Nicolas
David Brown· Nov 27, 2007, 21:10 UTC · re: Nicolas Pitre · lore

Re: Removing old data without disturbing tree?

On Tue, Nov 27, 2007 at 03:06:45PM -0500, Nicolas Pitre wrote:
Show 18 quoted lines
>On Tue, 27 Nov 2007, David Brown wrote:
>
>> An upstream tree I'm mirroring with git-p4 has decided to start checking
>> in large tarballs (150MB) periodically.  It's basically a prebuild version
>> of some firmware needed to run the rest of the software.
>> 
>> Git doesn't seem to have any problem with these tarballs (and is using a
>> lot less space than P4), but I have a feeling we might start running into
>> problems when things get real big.  Does anyone have experience with packs
>> growing beyong several GB?
>
>It should just work.  It was tested with artificial data sets but that's 
>about it.
>
>Now if those tarballs are actually multiple revisions of the same 
>package, you might consider storing them uncompressed and let Git delta 
>compress them against each other which will produce an even more 
>significant space saving.

I did manage to talk them into leaving them uncompressed. But, they are large, and don't seem to delta compress all that well. Maybe as more come, the compression will be better.

I guess this will be a good test case... It will probably take months or even a year or so for the repo to get up to several GB.

David

← back to recent threads