Mike,
If your nodes have more than one network interface it can happens
that we do not select the right one. There is a simple way to insure
that this does not happens. Create a directory named .openmpi in your
home area. In this directory edit the file mca-params.conf. This file
is loaded by Open MPI every time you start a job and it contain all
the MCA parameters you want/need to set.
For TCP you can get the list of available MCA parameters using
"ompi_info --param btl tcp". The one involved in selecting the
network are:
btl_tcp_if_include
btl_tcp_if_exclude
You just have to set one of them as they are exclusive. So if you
want to select the second network interface (let say it is named
eth1) you have to add in the mca-params.conf file the following line:
bl_tcp_if_include=eth1
Hope it helps,
george.
On Oct 31, 2005, at 10:46 AM, Mike Houston wrote:
We can't seem to run across TCP. We did a default 'configure'.
Shared
memory seems to work, but trying tcp give us:
[0,1,1][btl_tcp_endpoint.c:557:mca_btl_tcp_endpoint_complete_connect]
connect() failed with errno=113
I'm assuming that the tcp backend is the most thoroughly tested, so I
thought I'd ask in case we are doing something silly. The above is
caused when running the OSU NBCL mpi_bandwidth test.
Thanks!
-Mike
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
"Half of what I say is meaningless; but I say it so that the other
half may reach you"
Kahlil Gibran
