very common kernel modules slow down the boot process

[Karsten Hopp]

seth vidal schrieb:
> On Mon, 2008-04-07 at 23:33 -0400, Jesse Keating wrote:
>> On Mon, 2008-04-07 at 18:43 -0400, Alan Cox wrote:
>>> My comments were entirely accurate. I think you owe me an apology on list.
>>> To complain about be being inaccurate when you haven't actually *tried*
>>> the things and I have is most peculiar behaviour.
>> Just for funsies I attempted a text mode install in an i386 KVM guest
>> with 128 megs of ram.  Using today's rawhide boot.iso (which has stage2
>> on it) and pointing it at an http mirror.  Guess what, the install
>> completed fine, and booted after the fact.  I picked a minimal package
>> set from the package selector (included vim-enhanced though).  That's
>> still 428~ packages.
>>
>> So why do you have to go through gyrations to install Fedora on 128 meg
>> machines?  Or why do you have to install something other than Fedora?
>>
> 
> As an extra note. I tested yum's memory footprint for 3.2.14 vs 3.2.13
> and it should shave down the memory use even further.
> 
> On my i686 box running:
> yum -c my-rawhide.conf --installroot=/wherever install yum
> 
> Takes 42M of ram before the transaction is run.
> 
> During the transaction the memory use tops out at 51M
> 
> Now, granted that's only 70 pkgs for yum install yum on an empty chroot.
> 
> Still, pretty good.
> 
> -sv
> 

It's better than 3.2.13 on my x86_64 machine with > 3500 packages to update, although the
memory requirements are still quite high and the amount of time used for the transaction test is
extremely long (not yum's problem, that seems to be rpmlib).

19048 root      20   0 3475m 3.2g 9556 R 100.6 41.1  79:00.21 /usr/bin/python /usr/bin/yum update

Well, the yum traceback after 81 minutes is another matter, I've opened a bugzilla for that ;-(

    Karsten