Re: Fedora Release Engineering meeting summary for 2010-02-01

[Jesse Keating]

On Tue, 2010-02-02 at 07:54 -0800, John Reiser wrote:
>  >    * Second releng compose system was setup for testing 
> (Oxf13,
>  >      18:41:01)
>  >    * ACTION: Oxf13 will test multiple concurrent composes to compare
>  >      completion time  (Oxf13, 18:41:32)
>  
>  Related to performance of composing install media using pungi:
>  
>  https://bugzilla.redhat.com/show_bug.cgi?id=559647
>      quadratic work: genisoimage per .rpm on split .iso
>  
>  https://bugzilla.redhat.com/show_bug.cgi?id=560795
>      document practical usage  [such as "export TMPDIR=/dev/shm/pungi-$$"]
>  
>  --  
Wrong compose time.  The compose time we're talking about is from mash,
which is heavily disk bound.  In the full mash+pungi compose, mash takes
on the order of 6 hours, pungi takes on the order of 10 minutes.  These
composes don't even hit the code path you're complaining about.

-- 
Jesse Keating
Fedora -- Freedom² is a feature!
identi.ca: http://identi.ca/jkeating