YellowDog SMP kernel vs UP

[Tim Seufert]

On Monday, August 25, 2003, at 03:14  PM, Hassan Halta wrote:

> Hi all,
>
> We just started using YD for our MAC cluster. The cluster is a 1G Dual 
> CPU
> G4s. We were originally running MAC OS X on these Xserves, but recently
> moved to YD. The controller for this cluster, which is an identical
> machines to all of the Xserves, keeps hanging whenever we are using the
> SMP kernel. Basically the terminal won't be responsive and the whole
> cluster would be down. The controller does 1 thing that none of the 
> other
> Xserves do which is IP forwarding.
>
> Now, we tried running under UP kernel, and things just worked fine for 
> a
> longer time, but eventually it will crash again.
>
> I was wondering if there's a patch or a known problem for this.

Have you tested for hardware problems?  I'd swap one of the cluster 
computers for the controller computer in order to see if the problem 
follows the hardware.  This should be pretty easy to do with XServes: 
you can just swap the hard drive modules between the controller and one 
of the nodes.  Unless there's a dependency on Ethernet MAC addresses in 
the cluster software, no reconfiguration should be needed.