Re: [OMPI users] question about running openmpi with different interconnects

[Jeff Squyres]

On Apr 4, 2011, at 10:30 AM, Borenstein, Bernard S wrote:

> We have added clusters with different interconnects and decided to build one 
> OPENMPI 1.4.3 version to handle all the possible interconnects
> and run everywhere.  I have two questions about this :
>  
> 1 – is there a way for Openmpi to print out the interconnect it selected to 
> use at run time?  I am asking for an option similar to the –prot argument in 
> hpmpi/platform mpi to print the interconnect selected.  If this is not 
> implemented, I would like to suggest it as an enhancement.

Unfortunately, it is not implemented.  There's a long-standing ticket 
requesting this feature -- it's unfortunately not a simple task, since Open MPI 
opens connections lazily.

https://svn.open-mpi.org/trac/ompi/ticket/1207

> 2 – I have built Openmpi to allow tcp, mx, gm and ib.  When running on a tcp 
> only cluster and specifying –mca btl tcp,sm,self, I get errors like this
> [erb426:08967] Error in mx_init (error No MX device entry in /dev.)

Try also specifying:

   mpirun --mca pml ob1 ...

See the README for details of the "ob1" vs. "cm" PMLs.

-- 
Jeff Squyres
jsquy...@cisco.com
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/

[OMPI users] question about running openmpi with different interconnects

[Borenstein, Bernard S]

We have added clusters with different interconnects and decided to build one 
OPENMPI 1.4.3 version to handle all the possible interconnects
and run everywhere.  I have two questions about this :

1 - is there a way for Openmpi to print out the interconnect it selected to use 
at run time?  I am asking for an option similar to the -prot argument in 
hpmpi/platform mpi to print the interconnect selected.  If this is not 
implemented, I would like to suggest it as an enhancement.

2 - I have built Openmpi to allow tcp, mx, gm and ib.  When running on a tcp 
only cluster and specifying -mca btl tcp,sm,self, I get errors like this
[erb426:08967] Error in mx_init (error No MX device entry in /dev.)

Since I am calling out the interconnect to use, why would I be getting an error 
like this??

Bernie Borenstein
The Boeing Company