Problem Booting the Slaves

Dave Leimbach dave at MPI-Softtech.Com
Fri Dec 15 01:52:12 PST 2000


----- Original Message -----
From: "Erik Arjan Hendriks" <erik at hendriks.cx>
To: "Dave Leimbach" <dleimbac at mpi-softtech.com>
Cc: "David Leunen" <leunen.d at fsagx.ac.be>; <beowulf at beowulf.org>;
<beowulf-newbie at fecundswamp.net>
Sent: Friday, December 15, 2000 9:47 AM
Subject: Re: Problem Booting the Slaves


>
> "Unable to handle kernel paging request" just means that there was a
> page fault that the kernel didn't know what to do with.  That's
> usually caused by referencing a bad pointer.  You can also do it by
> branching to oblivion.  Judging by the addresses, it looks like the
> bad pointer case.
So a kernel driver/module/function that dereferences a pointer to NULL
would cause the same error?   That sounds about right.

Also its interesting to note that this happened during the "two kernel
monte" way
of loading a different linux kernel from an already running linux kernel
without
a reboot.

There is some pretty interesting stuff going on in there in the two kernel
monte.
All of which seems pretty well documented on Scyld's website!
:)

Dave

>
> That being said... 16MB is not enough to boot a Scyld beowulf slave.
> I believe the minimum is somewhere around 64MB.  There are some rather
> large ram disks involved.  You need this much RAM to get things off
> the ground regardless of whether or not you're trying to do a diskless
> system.
>
> In a sense, all systems start diskless.  Most of the stuff that's in
> these ram disks gets freed when node setup is finished.  The setup
> scripts free it right before setting the node state to "up".
>
> Scyld's next release will be a lot better for low memory systems.
>
> > The one that comes up unavailable has 256MB of RAM which is plenty of
RAM
> > for the ramdisks.  Scyld has an option in the beowulf distribution to
run
> > diskless.  I assume that this requires a large ramdisk that you can't
get
> > on the 16MB node.
> >
> > Dave
> >
> > On Fri, 15 Dec 2000, David Leunen wrote:
> >
> > > Hello,
> > >
> > > With BeoSetup's Scyld, I assigned the nodes to the cluster by draging
> > > the ether-addresses, and cliked apply. Then I did
> > >
> > > "bpctl -S all -s reboot"
> > >
> > > but nothing appened: the states of the nodes stayed "down". On the
> > > slaves, the output was:
> > >
> > > ...
> > > monte: command line: panic=30 ramdisk_size=131072 apm=power-off
> > > monte: Loading 7 sectors of setup code
> > > monte: Loading kernel data at0x100000
> > > monte: Loaded 642144 bytes of kernel data
> > > Unable to handle kernel paging request at virtual address 9f4d4dcb
> > > current->tss.cr3=00195000, %cr3=00195000
> > > *pde=00000000
> > > Oops: 0000
> > > CPU:  0
> > > EIP:  0010:[<c1800060>]
> > > EFLAGS: 00010282
> > > ...
> > >
> > > after that, there are all the register and stack status, and it is
> > > frozen. The slaves are P200 16Ram Diskless. I tried with a double PII
> > > 256Ram Diskless and it turn 'unavailable' after reboot. What is wrong
> > > with the other ones?
> > >
> > > Thanks.
> > >
> > > David
> > >
> > > _______________________________________________
> > > Beowulf mailing list
> > > Beowulf at beowulf.org
> > > http://www.beowulf.org/mailman/listinfo/beowulf
> > >
> >
> >
> > _______________________________________________
> > Beowulf mailing list
> > Beowulf at beowulf.org
> > http://www.beowulf.org/mailman/listinfo/beowulf
>
> --
> Erik Arjan Hendriks          Printed On 100 Percent Recycled Electrons
> erik at hendriks.cx                   Contents may settle during shipment
>






More information about the Beowulf mailing list