Thanks.
I think I figured out the problem. I found that in my .ssh/
known_hosts there were several "bad" keys associated with some of the
machines in the gridengine pool. My hypothesis is that when mpirun
was establishing the connection topology of the processes there was
some process pairs that failed to complete the connection due to the
bad ssh keys. I don't have explicit evidence for this since there
was no ssh error output generated.
I generated new keys for all the amd64 machines in the gridengine
pool for which there was an offending key. Now my job runs with a
set of machines that includes ones that had previously failed. I
will assume for now that the problem is fixed.
~Tim
On Sep 13, 2007, at 12:06 PM, Pak Lui wrote:
Hi Tim,
You could try setting -mca pls_gridengine_verbose 1 to show whether
SGE
is able to start the ORTE daemons on the remote nodes successfully.
It seems you are having the problem previously asked by another user,
Perhaps you may want to follow this thread and check your ifconfig
settings to see if anything suspicious?
http://www.open-mpi.org/community/lists/users/2007/02/2669.php
My 2 cents...
Tim Campbell wrote:
Greetings,
I am using OpenMPI v1.2.3 via SGE on a network of amd64
workstations. When mpirun tries to start the processes on certain
nodes I get the following error output.
[sr70][0,1,2][btl_tcp_endpoint.c:
572:mca_btl_tcp_endpoint_complete_connect] connect() failed with
errno=111
[sr71][0,1,3][btl_tcp_endpoint.c:
572:mca_btl_tcp_endpoint_complete_connect] connect() failed with
errno=111
Using perl -e 'die$!=111' I see that the error message is "Connection
refused". I am able to connect to both nodes in question via ssh
and/
or rsh. I changed btl_base_debug to 2, but that did not provide
additional information.
What are some possible issues that might be causing this? What can I
do to get more information?
Thanks,
~Tim
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
--
- Pak Lui
pak....@sun.com
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
