From: Kevin R. <Kev...@dr...> - 2004-06-08 20:43:31
|
Thanks for the good work. I've managed to boot a 9 node cluster under the 2.6.5 kernel and bproc-4.0.0pre4 using PXE and etherboot. (Insure that you do NOT "Use register arguments (EXPERIMENTAL)" CONFIG_REGPARM under the "Processor type and features" in the kernel configuration). This was noted by Steven James in an earlier posting (thanks). I didn't have much success with beoboot. The node would successfully download the stage2 image, but then immediately reset and reboot the node. However I'm fairly happy with booting via PXE (and in one node's case; etherboot). Now for the question: How long before lam-mpi catches up with the API changes between bproc-3.2.6 and bproc-4.0.0pre4? A quick attempt at building lam-7.1a1r9667 (a nightly snapshot) indicates that lam-mpi needs to change the way it usings the following funtions: bproc_nodestatus bproc_getnodebyname bproc_nodenumber I suspect that I could do this, but I thought I'd check to see if anyone else has already tackled this issue. Thanks, Kevin -- Kevin Russell (DND/DRDC/MES/TDG) Kev...@dr... (403) 544-4746, DRDC Suffield, Box 4000, Medicine Hat, AB, Canada, T1A 8K6 |