Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-16 Thread Ricardo Fernández-Perea
They are just as you say but while one run until the end (case 3 using --deamons-debug) the other hangs (Case 1) in the case 1 even if is only in one node with plm_rsh_no_tree_spawn 1 flag (that as you say shouldn't do anything), the process hangs, while in the case 2 without this flags in the

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-15 Thread Ralph Castain
Forgive me, but I am now fully confused - case 1 and case 3 appear identical to me, except for the debug-daemons flag on case 3. On Jul 15, 2014, at 7:56 AM, Ricardo Fernández-Perea wrote: > What I mean with "another mpi process". > I have 4 nodes where there is process that use mpi and whe

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-15 Thread Ricardo Fernández-Perea
What I mean with "another mpi process". I have 4 nodes where there is process that use mpi and where initiated using mpirun from the control node already running when I run the command against any of those nodes it execute but when I do it against any other node it fails if no_tree_spawn flag

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-15 Thread Ralph Castain
I'm afraid I don't understand your comment about "another mpi process". Looking at your output, it would appear that there is something going on with host nexus17. In both cases, mpirun is launching a single daemon onto only one other node - the only difference was in the node being used. The "n

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-15 Thread Ricardo Fernández-Perea
I have try if another mpi process is running in the node already the process run $ricardo$ /opt/openmpi/bin/mpirun --mca plm_rsh_no_tree_spawn 1 -mca plm_base_verbose 10 -host nexus16 ompi_info [nexus10.nlroc:27397] mca: base: components_register: registering plm components [nexus10.nlroc:27397]

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-14 Thread Ralph Castain
Hmmm...no, it worked just fine for me. It sounds like something else is going on. Try configuring OMPI with --enable-debug, and then add -mca plm_base_verbose 10 to get a better sense of what is going on. On Jul 14, 2014, at 10:27 AM, Ralph Castain wrote: > I confess I haven't tested no_tree

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-14 Thread Ralph Castain
I confess I haven't tested no_tree_spawn in ages, so it is quite possible it has suffered bit rot. I can try to take a look at it in a bit On Jul 14, 2014, at 10:13 AM, Ricardo Fernández-Perea wrote: > Thank you for the fast answer > > While that resolve my problem with cross ssh authentica

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-14 Thread Ricardo Fernández-Perea
Thank you for the fast answer While that resolve my problem with cross ssh authentication a command as /opt/openmpi/bin/mpirun --mca mtl mx --mca pml cm --mca plm_rsh_no_tree_spawn 1 -hostfile hostfile ompi_info just hung with no output and although there is a ssh connexion no orte program is

Re: [OMPI users] Can't run with more than two nodes in the hostfile

2014-07-14 Thread Ralph Castain
During the 1.7 series and for all follow-on series, OMPI changed to a mode where it launches a daemon on all allocated nodes at the startup of mpirun. This allows us to determine the hardware topology of the nodes and take that into account when mapping. You can override that behavior by either