I pulled the node, checked everything, cleaned ram and card connections, and re-installed it. PXE booted with kvm attached and watched virtual consoles. It is indeed doing other stuff on the other virtual consoles, but nothing significant is reported. Also, I let it sit all night, came back a few minutes ago, and it was stuck where I left it last night. I just replaced that node. I'll try taking it totally apart and see what I can find.

- Don

----- Original Message -----
From: Troy Dawson <[log in to unmask]>
Date: Monday, October 23, 2006 3:53 am
Subject: Re: freeing kernel memory
To: Donald E Tripp <[log in to unmask]>
Cc: [log in to unmask]

> Stephen John Smoogen wrote:
> > On 10/21/06, Donald E Tripp <[log in to unmask]> wrote:
> >> I have a supermicro board based server that freezes at "freeing 
> unused 
> >> kernel memory" on SL43 using pxeboot. Any ideas?
> > 
> > Make sure the system BIOS is the latest stable version for that 
> > motherboard.
> > 
> > Have the boot lines try noapic or apci=off. If those work then there
> > is probably a motherboard issue with the IRQ routing causing ti to
> > crash out.
> > 
> >>
> >> - Don
> >>
> > 
> > 
> 
> Also, give it some time.
> I've had several people say that their machine freezes at this 
> point, 
> and usually the computer is doing something in the background.  
> After 
> 10-15 minutes it returns.  It usually is only a one time thing also.
> What is it doing?
> Well, the most common thing is that the output is being redirected 
> to 
> the serial console.  At that point "freeing unused kernel memory" 
> is 
> when *everything* is being re-directed to the serial console so 
> you 
> don't see anything until the machine is completely up and you get 
> a 
> console prompt.
> But when it's taking so long (more than 30 seconds), then people 
> think 
> the machine is frozen, they reboot it, which then means it's going 
> to do 
> a scan of your disks, which takes an even longer time, so then 
> they sit 
> there for 5 minutes, and think, oh, it's frozen, reboot again, and 
> the 
> cycle continutes.
> It's usually when they see it frozen, give up, go to lunch, come 
> back, 
> and then magically everything is fixed and they don't know why.
> (As a side note, this most often happens when people select 
> *install 
> everything*)
> 
> Now true, there might be something actually hanging, and the 
> advice that 
> Stephen gave will actually work.  That's often what the problem is 
> for 
> the time's there really is a problem.
> 
> Troy
> -- 
> __________________________________________________
> Troy Dawson  [log in to unmask]  (630)840-6468
> Fermilab  ComputingDivision/CSS  CSI Group
> __________________________________________________
>