My company has received an evaluation unit of the new Nehalem-based Dell R610 1U server. It has 12G of RAM and plenty of disk space. Two 2.26 Quad-core CPUs, yielding 16 cores seen by the OS (Simultaneous Multi-Threading). It also has 4 built-in NICs and a e1000 two-port NIC card, giving me 6 total NICs.
I deploy this node from a Dell 1435 frontend, running Platofrm OCS 5.1 Dell Edition, as a diskless node, and everything installs fine, and I'm happy.
I try to deploy this node as a diskful compute node, and it doesn't work. Anaconda is loading the e1000 driver before it loads the Broadcom driver, thus transfering eth0 to the add-on NIC card.
When it does this, the installer getsa loaded onto the compute node, then loads up the two drivers in the wrong order and then it cannot find the install server anymore.
While I understand that the workaround for me right now is to remove the add-on NIC, this issue may crop up in the future for other users. Is there any way of actually correcting this without customizing the install scripts?


LinkBack URL
About LinkBacks
Reply With Quote
