Hi guys,<br>
<br>
&nbsp; I seem to have encountered an error while trying to run an MPMD
executable through Open MPI&#39;s &#39;-app&#39; option, and I&#39;m wondering if
anyone else has seen this or can verify this?&nbsp; <br>
<br>
Backing up to a simple example, running a &quot;hello world&quot; executable
(hwc.exe) works fine when run as:&nbsp; (using an interactive PBS session
with -l nodes=2:ppn=4)<br>
&nbsp;mpiexec -v -d&nbsp; -machinefile $PBS_NODEFILE -mca oob_tcp_if_exclude eth0 -mca pls_rsh_agent ssh -np 8 ./hwc.exe <br>
<br>
But when I run what should be the same thing via an &#39;--app&#39; file (or implied command line) liks the following fails:<br>
&nbsp;mpiexec -v -d&nbsp; -machinefile $PBS_NODEFILE -mca oob_tcp_if_exclude eth0
-mca pls_rsh_agent ssh&nbsp; -np 6 ./hwc.exe : -np 2 ./hwc.exe<br>
<br>
&nbsp; My understanding is that these are equivalent, no?&nbsp; But the latter
example fails with multiple &quot;Software caused connection abort (103)&quot;
errors, such as the following:<br>
[xxx:13978] [0,0,2]-[0,0,0] mca_oob_tcp_peer_try_connect: connect to
xx.x.2.81:34103 failed: Software caused connection abort (103)<br>
<br>
&nbsp; Any thoughts?&nbsp; I haven&#39;t dug around the source yet since this could
be a weird problem with the system I&#39;m using.&nbsp; For the record, this is
with OpenMPI 1.2.4 compiled with PGI 7.1-2.<br>
<br>
&nbsp; As an aside, the &#39;-app&#39; syntax DOES work fine when all copies are
running on the same node.. for example, having requested 4 CPUs per
node, if I run &quot;-np 2 ./hwc.exe : -np 2 ./hwc.exe&quot;, it works fine.&nbsp; And I did also try duplicating the mca parameters after the colon since I figured they might not propagate, thus perhaps it was trying to use the wrong interface, but that didn&#39;t help either.
<br>
<br>
&nbsp; Thanks very much,<br>
&nbsp; - Brian<br>
<br>
<br>
Brian Dobbins<br>
Yale University HPC<br>