Re: [Internet]Re: [PATCH] send-pack.c: add config push.useBitmaps

["kylezhao(赵柯宇)" <kylezhao@tencent.com>]

> > I'm glad that this references `pack.useBitmaps`. I wonder if that config is sufficient for your purposes: do you expect to use your bitmaps to generate pack-files in any other way > > I just want to be sure that using pack.useBitmaps=false would not be 
> > sufficient for your situation (and probably most situations).
> 
> I think the only other affected scenario on the client side would be repacking. And in theory most clients are repacking in the background anyways, so any speed-ups from using bitmaps wouldn't be noticeable anyway.
> 
> I think just relying on the existing pack.useBitmaps config should be sufficient here.

In fact ,we also use "git push" on our server side. 
Each git repositories have multiple replicas on our servers to improve system disaster tolerance and read performance.

For example, a git repo will be distributed on multiple servers (like server-1, server-2, server-3).
If user pushes the pack to server-1, then server-1 will call "git push" to transfer the objects data to server-2 and server-3.
And users can clone from all the server mentioned above.
Under such a process, our system works well most of the time.

If we set pack.useBitmaps=false, "git upload-pack" will also be affected.
For my situation, it would be sufficient when set both pack.useBitmaps=true and push.useBitmaps=false.

> But I suspect that in this case the bitmaps are just simply stale, and
> that a "git repack -adb" or more aggressive background maintenance would
> make things quite a bit faster.

It doesn't seem to be the reason.
I have already called  "git repack -adb" in this case[1] but that didn't seem to fix the push performance issue.
You can see that the git repo had only one pack at that time.

[1] https://lore.kernel.org/git/b940e705fbe9454685757f2e3055e2ce@tencent.com/

Thanks,
- Kyle