On 01:45 Sat 24 Jan , Robin H. Johnson wrote:
> Secondary delay is that I asked for some input on hardware and tuning re
> managing a very large repo, and still haven't got a response from my
> contact (at kernel.org). Regardless of the tuning, we're going to need
> to figure out a box with lots of RAM for the repo size.

19:56 <  warthog9 > dberkholz- git is basically unnoticable by itself
19:58 < dberkholz > warthog9: so RAM and IO are both pretty minimal for git 
                    itself?
19:59 <  warthog9 > dberkholz- we were running git+gitweb+all of our http 
                    mirroring on one box for a long time, and it was gitweb 
                    that was killing us
20:00 <  warthog9 > dberkholz- to give you an idea, the machines that are 
                    dealing with our pub mirroring (I.E. www.kernel.org/pub/) 
                    gitweb (android.git, mirror.git, git), git, etc is at a 
                    wopping load of 1
20:02 <  warthog9 > (mind you this is a dual quad core with 16G of ram and like 
                    4T of sas storage but yeah)
20:03 <  Ramereth@> warthog9: do you do anything special w/ xinetd or tmpfs w/ 
                    the git-daemon?
20:03 <  warthog9 > Ramereth- nope
20:04 <  warthog9 > I run different git-daemons for things like mirror.git, 
                    android.git and git but that's because they are on separate 
                    ips

-- 
Thanks,
Donnie

Donnie Berkholz
Developer, Gentoo Linux
Blog: http://dberkholz.wordpress.com