Slave node problem

Sean Dilda agrajag at scyld.com
Tue Aug 21 01:18:36 EDT 2001


On Mon, 20 Aug 2001, ericf at whispers.org wrote:

> Hi, I just recently installed the Scyld Beowulf software (27bz-7) and have
> ran into some difficulty with slave nodes.  The front end setup went
> smoothly and I made my floppy disks for the slaves.  When I boot up the
> slaves, they go through the boot process then start the RARP sequence, so
> I dragged their MAC adresses from unknown to configured nodes and hit
> apply.  The slave nodes appear to get an IP address then (and this happens
> pretty fast) the machine reboots.  I noted it said neighbour table
> overflow right after the IP assignment.  I've waded thru the mailing list
> archives and the only remotely relevent thing I found to my problem didn't
> really have an answer.  Any help would be greatly appreciated.

What happened after the machine rebooted?

In a normal run of events, once you hit the apply button, then slave
node will get an IP, then imediately request a phase2 beoboot image from
the master and download that.  This can happen pretty quickly over a
fast network.  Once it has the phase2 beoboot image, it will do what we
call the 2-kernel monte.  It will boot the kernel that is on the phase2
floppy.  You will see the normal stuff for a kernel boot, however this
isn't quite a normal reboot as it never goes back to BIOS or anything
like that.  Once it has booted the kernel from the phase2 image, it will
RARP again, and once it has its IP it will start up sendstats and
bpslave, and once bpslave is up, the status of the slave node should
change in beosetup.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
URL: <http://www.clustermonkey.net/pipermail/beowulf/attachments/20010821/86743a6f/attachment.sig>


More information about the Beowulf mailing list