Menu
▾
▴
Re: [BProc] boot.img never comes on clustermatic
|
From: Florent C. <Flo...@un...> - 2003-01-24 17:50:50
|
thank you very much Lothar.
I do not think your patch
helped, because this is how I managed to make Clustermatic work :
since we are building a diskless cluster at the same time,
and I was destroying floppy disks in my tries to make a
working one, and moreover I got tired of waiting 2 minutes at each boot
try, I decided to give a try to PXElinux. After some problems
(the switch allowed ethernet ports to the nodes slower
than PXE/dhcp would give up - thanks to the very speed
in boot I was looking for - had to solve it
bu setting spanning tree to fast and channel to
off, as suggested by Cisco), I managed to load directly
a phase 2 image by pxe. So everything seems to work ...
thanks anyway
ot...@tr... wrote:
> I copy down there two messages which made it work..though I am not
> totally sure it's
> your problem
>
> Lothar
>
>
> I just saw something like this on a cluster here and I fixed it. If
> you're having the same problem, the following patch should fix it for
> you. This only modifies the "beoserv" binary (which runs on the
> master and does things like server boot images and RARP responses) so
> you won't have to change the slave boot images or anything.
>
> Index: mcsend.c
> ===================================================================
> RCS file: /users/hendriks/repository/beoboot/mcsend.c,v
> retrieving revision 1.25
> retrieving revision 1.26
> diff -u -r1.25 -r1.26
> --- mcsend.c 27 Aug 2002 16:25:13 -0000 1.25
> +++ mcsend.c 17 Dec 2002 21:49:13 -0000 1.26
> @@ -28,7 +28,7 @@
> * negligence or otherwise) arising in any way out of the use of this
> * software, even if advised of the possibility of such damage.
> *
> - * $Id: mcsend.c,v 1.25 2002/08/27 16:25:13 hendriks Exp $
> + * $Id: mcsend.c,v 1.26 2002/12/17 21:49:13 hendriks Exp $
> *--------------------------------------------------------------------*/
> #include <sys/time.h>
> #include <sys/types.h> @@ -1029,6 +1029,10 @@
> }
> break;
> case SND_TIME_WAIT:
> + if (ifc->sendok > 0 &&
> + !FD_ISSET(ifc->fd, wset) && sender_ready(s))
> + FD_SET(ifc->fd, wset);
> +
> timeleft = SENDER_TIMEOUT - (now.tv_sec - s->lastuse);
> if (timeleft <= 0) {
> sender_discard(s);
>
>
>
> you'll have to rebuilt beoboot. Grab the source RPM (included in
> Clustermatic 3) and apply this patch to it.
>
> You'll have to do something like:
>
> rpm -i beoboot-....src.rpm
>
> rpmbuild -bp /usr/src/redhat/SPECS/beoboot.spec
>
> cd /usr/src/redhat/BUILD/beoboot-....
>
> patch -p1 < patchfile
>
> make beoserv
>
>
> Then replace the beoserv in /usr/sbin with the one built there. See
> local Linux guru for more help on building stuff. :)
>
> - Erik
>
>> Erik A. Hendriks wrote:
>>
>> >On Mon, Dec 16, 2002 at 10:05:11AM -0800, lo...@tr... wrote:
>> > >
>> >>Well that's how it goes. Looks to me as if the problem is
>> >>on the master side....but no idea what.
>> >>
>
>
>
>
> Lothar
>
>
> Florent Calvayrac wrote:
>
>> Hi
>>
>> I gave a try to clustermatic today, on a test cluster
>> made of 8 nodes with dual pentium 3 processors,
>> serverwork chipset with integrated eepro100 , + one eepro
>> on the master, myrinet 2000 boards and switch.
>>
>> Either with fast ethernet or myrinet the nodes
>> hang on the boot.img fetch after they got an
>> IP from RARP. I tried the mcastbcast hack,
>> crossover cables instead of our Cisco switch,
>> and boot over myrinet (with different MACs ) :
>> all the same, nothing happens
>> and tcpdump does not show anything after the RARP
>> resolution.
>>
>> I chose addresses in the 192.168.33 range in order
>> not to interfer with the campus here.
>>
>> Did I forget something in the redhat 8.0 security
>> settings ? I set no firewall and started all services...
>>
>> Please help
>>
>> thanks in advance
>>
>>
>
>
--
Florent Calvayrac | Tel : 02 43 83 26 26
Laboratoire de Physique de l'Etat Condense | Fax : 02 43 83 35 18
UMR-CNRS 6087 | http://www.univ-lemans.fr/~fcalvay
Universite du Maine-Faculte des Sciences |
72085 Le Mans Cedex 9
|
