unable to boot machine with tesla card

[wcochran]

I am trying to set up a cluster of 16 boxes connected with InfiniBand which each have a Tesla card (M2075). There is also another "head" machine that does *not* have a Tesla card.

The machines are branded SuperMicro, model number sys_1026GT-TF
Contain Intel Xeon X5679 3.2Ghz.

(A) What is the preferred Linux distro for cluster machines using nVidia Tesla cards and MPI?

(B) The machines came from the vender with Linux Suse. I have no problem installing Linux on the head machine. I have tried installing Ubuntu 11.10 and CentOS 6.2 (x84_64), but the boot kernels hang.

Here is a screen shot of the Ubuntu install kernel at hang:


Here is a screen shot of the CentOS install kernel at hang:


Any idea why the boot kernels hang when a Tesla card is present?
Ugh....

[Plagman]

Nouveau panicks your kernel when trying to initialize this card; please remove nouveau from your system and install the NVIDIA driver.

[wcochran]

Originally Posted by Plagman

Nouveau panicks your kernel when trying to initialize this card; please remove nouveau from your system and install the NVIDIA driver.

This is the boot DVD image... I haven't even got to the point where I can install a driver...

BTW, the CentOS boot install works fine... the modprobe process gets a kill -9 when trying to recognize the tesla card, but installation proceeds as normal. Of course, I don't know if this is a preferred distro for doing this kind of thing.

[Plagman]

Try a text mode installer, I know Ubuntu provides them as "alternate images".