[Mitchelle Rasquinha]

How is everyone running their jobs on the cluster. Dont u have to start the mpd deamon on each node??

[Dustin]

You can start with using SMP, since each machine has 8 cores. I think
the MPICH2 binaries provided can only use shared memory.

By the way, someone is hogging the ellington.cc node. The PBS system
is not working properly; the armstrong.cc and whiteman.cc nodes are
not being provisioned. I already reported this to Rich and Chad.

Cheers,
Dustin

[Dustin]

I lied, nemesis can use sockets too. Check out section A.1 of the
install guide for a tutorial on how getting this running. But don't
hog nodes, or someone may kill -9 you.
http://www.mcs.anl.gov/research/projects/mpich2/documentation/files/mpich2-1.0.8-installguide.pdf

[Dustin]

Please kill your processes that have each used more than 1 hour of CPU time:
20621 mrasquin 25 0 67504 1960 1744 R 100 0.0 71:04.62 mysolver
20622 mrasquin 25 0 67500 1956 1744 R 100 0.0 71:18.72 mysolver
20620 mrasquin 25 0 67500 1960 1744 R 100 0.0 71:21.07 mysolver
20623 mrasquin 25 0 67504 1984 1768 R 100 0.0 71:11.32 mysolver
20624 mrasquin 25 0 67500 1956 1744 R 100 0.0 71:19.11 mysolver
20619 mrasquin 25 0 67500 1956 1740 R 100 0.0 71:24.85 mysolver

[Mitchelle Rasquinha]

i did kill them... i cant seem to log into any of the other nodes and the jobs with more than 6 nodes take forever to complete.

[Mitchelle Rasquinha]

have u tried using mpi get and put... i cant seem to make that work

On Fri, Dec 5, 2008 at 8:08 PM, Dustin <publi...@dustin.li> wrote: