[MPICH] mpirun issue on 64bit Xeon

Mike Zupan hijinks at gmail.com
Wed Aug 31 06:27:48 CDT 2005


I think i might have figured out the issue for my mpirun problem. I setup a 
testbed 2 node cluster on 32bit p4's and everything worked. I do the same 
steps on my 64bit system and it doesn't work.. I add the master node of the 
64bit system as a slave on my 32 bit system and this happens 

wind at homer:~/mpich-1.2.7/examples/basic$ ~/mpich-1.2.7/bin/mpirun -np 2 
cpilog
Process 0 running on homer.meso.com <http://homer.meso.com>
Process 1 running on unused.ip.244.meso.com <http://unused.ip.244.meso.com>
pi is approximately 3.1415926535899401, Error is 0.0000000000001470
wall clock time = 0.851333

that is without the 64bit system

here it is with it

wind at homer:~/mpich-1.2.7/examples/basic$ ~/mpich-1.2.7/bin/mpirun -np 3 
cpilog
rm_23376: p4_error: Could not gethostbyname for host homer; may be invalid 
name
: 61
p0_5892: p4_error: Child process exited while making connection to remote 
process on cluster1: 0
Killed by signal 2.
p0_5892: (62.935995) net_send: could not write to fd=4, errno = 32


mpich is compiled seperately on the 64bit system.. right now no nfs mounts 
or sharing home directories.. this is all on centos 4.0 systems.. basic 
installs

-- 
http://www.zcentric.com/wiki
My Linux Howto/Tips Wiki
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20050831/6d16d1b8/attachment.htm>


More information about the mpich-discuss mailing list