[scyld-users] Beoboot failure

Donald Becker becker at scyld.com
Mon Jan 30 13:34:08 PST 2006

On Fri, 27 Jan 2006, Gardner Pomper wrote:

> I am trying to get my test Scyld cluster up. I have Scyld v29cz running on
> my head node, and I am trying to get one slave node booting from a CD. The
> slave finds the master, and is assigned an IP address, but it fails with the
> following error:

What version are you using?
Are you booting from the distribution CD, or an "iso image" created from 
Are you running on a x86_64 or IA32?
> Boot information received for interface eth0 from RARP:
>   Assigned IP address: /
>   Server:  boot file '/var/beowulf/boot.img'
>   BProc master port 2223,
>   VMA server port 1556, protocol tcp.
> boot: installing module "kmonte"
> connect: no route to host
> Boot image download failure: numerical result ouf of range

Hmmm, this appears to be a routing problem.

We have seen a similar problem once before in the quick-remaster case, 
where the old default route prevents a new route from being added, even 
though the interface was shut down.  (This is a kernel routing bug that 
can be worked around by explicitly deleting the old default route.)

> I hope I got all this right, because it automatically clears the screen and
> reboots after 5 seconds and I haven't found a way to stop it.

The parameter 'fatal_delay' should set that timeout, and the default is 
normally 120 seconds.  The only five second delay is in Two Kernel Monte, 
and that shouldn't be called.  (FYI: The TKM delay may changed with the 
'bootmode' flags.  The upper 8 bits specify the delay in seconds before 
switching kernels.  This is pretty much useful only for 
debugging and developers, and isn't likely the delay here.)

Donald Becker				becker at scyld.com
Scyld Software	 			Scyld Beowulf cluster systems
914 Bay Ridge Road, Suite 220		www.scyld.com
Annapolis MD 21403			410-990-9993

More information about the Scyld-users mailing list