I don't know whether this is a bug or my setting problem.
So I want to check this before reporting bug.
This is the error message from mpirun
[[22878,1],3][../../../../../../ompi/mca/btl/tcp/btl_tcp_endpoint.c:589:mca_btl_tcp_endpoint_start_connect] from ray003 to: gamma00 Unable to connect to the peer 10.51.111.1 on port 4: Network is unreachable
gamma00 has several different IP address with virtual interfaces.
eth2 = 10.51.1.1 - MPI network
eth2:0 = 10.51.2.1
eth2:1 = 10.51.111.1
Since ray003 is in 10.51.1.0 network, they should connect gamma00 using 10.51.1.1.
I also specified it in /etc/hosts in ray003.
This is an new thing for Squeeze, The same configuration was working with Lenny.
Another thing I found was that it didn't recognize the default host file at /etc/openmpi/openmpi-default-hostfile
It worked after I manually added the line
orte_default_hostfile = /etc/openmpi/openmpi-default-hostfile
in the /etc/openmpi/openmpi-mca-params.conf
Is there anybody have an answer for these?
-------------- Innovation for the Future Radiation Oncology (http://rophys.meds.case.edu)
(Samuel) Byeongjun Park, Ph.D. -- Research associate of Sohn Lab.
Case Western Reserve Univerisity, School of Medicine
Visiting: Wood building W517, Phone: 1-216-368-6583