Scyld module configuration
ben.delong at dnamerican.com
Wed Jul 18 12:01:52 PDT 2001
I've done some more checking into this problem. I added the line you
specified into the config.boot file and rebuilt the boot images for
phase 2. The failure occurs when the modules are loaded for phase 2 (no
suprise). It appears from the boot text that the modarg parameter is
being sent to the pci-scan module instead of the 3c59x module, pci-scan
fails to load because of the invalid argument, and therefore the 3c59x
module fails to load because it depends on pci-scan.
Is there a way that I can look at the modules dependancies generated by
beoboot after (during) boot image creation.
Thanks for your help.
Sean Dilda wrote:
>On Fri, 13 Jul 2001, Ben DeLong wrote:
>>A couple of weeks ago I posted a question regarding a specific 3Com
>>network card that wouldn't detect 100Mb/sec with the 2.2.x linux kernel
>>(Mandrake 7.2) and instead only ran at 10Mb/sec. I was able to fix the
>>problem by passing an option to the 3c59x module when it was loaded.
>> Specifically I added the line "options eth0 options=8,8" to the
>>Yesterday I got around to installing Scyld on our 4 node cluster and
>>bumped into the same problem - all the cards were running at 10Mb after
>>install. On the head node the problem was easy enough to fix by adding
>>the options line to the conf.modules file on the head. My question is
>>how do I pass this parameter on the slave nodes? I have looked through
>>all the documentation I have and I can't find an answer.
>In the /etc/beowulf/config.boot file you can add the line:
>modarg 3c59x options=8,8
>Then remake your phase1 and phase2 images and it should do what you
>want. This should be documented in the reference guide, available at
>By the way, what version of our software are you running? I ran into a
>problem with a 3com card not detecing 100M in our preview release, but
>the updated drivers in the 27bz-7 release worked without a problem.
More information about the Beowulf