Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

From: SLIM H.A. (h.a.slim_at_[hidden])
Date: 2007-07-06 07:37:42


Dear Michael
 
I have now tried both
 
mpirun --mca btl mx,sm -np 4 ./cpi
 
which gives the same error message again, and,
 
mpirun --mca btl mx,sm,self -np 4 ./cpi_gcc_ompi_mx
 
actually locks some of the mx ports but not all 4, ie this is the output
from endpoint info:
 
1 Myrinet board installed.
The MX driver is configured to support up to 4 endpoints on 4 boards.
===================================================================
Board #0:
Endpoint PID Command Info
<raw> 5061 mx_mapper
0 20315 cpi
There are currently 1 regular endpoint open
 
This is the output from the node:
>mpirun --mca btl mx,sm,self -np 4 ./cpi_gcc_ompi_mx
[node001:20312] mca_btl_mx_init: mx_open_endpoint() failed with
status=20
[node001:20314] mca_btl_mx_init: mx_open_endpoint() failed with
status=20
[node001:20313] mca_btl_mx_init: mx_open_endpoint() failed with
status=20

Thanks
 
Henk

 

________________________________

        From: users-bounces_at_[hidden]
[mailto:users-bounces_at_[hidden]] On Behalf Of Michael Edwards
        Sent: 05 July 2007 18:06
        To: Open MPI Users
        Subject: Re: [OMPI users] openmpi fails on mx endpoint busy
        
        
        If the machine is multi-processor you might want to add the sm
btl. That cleared up some similar problems for me, though I don't use
mx so your millage may vary.
        
        
        On 7/5/07, SLIM H.A. <h.a.slim_at_[hidden]> wrote:

                Hello
                
                I have compiled openmpi-1.2.3 with the
--with-mx=<directory>
                configuration and gcc compiler. On testing with 4-8
slots I get an error
                message, the mx ports are busy:
                
>mpirun --mca btl mx,self -np 4 ./cpi
                [node001:10071] mca_btl_mx_init: mx_open_endpoint()
failed with
                status=20
                [node001:10074] mca_btl_mx_init: mx_open_endpoint()
failed with
                status=20
                [node001:10073] mca_btl_mx_init: mx_open_endpoint()
failed with
                status=20
        
------------------------------------------------------------------------
                --
                Process 0.1.0 is unable to reach 0.1.1 for MPI
communication.
                If you specified the use of a BTL component, you may
have
                forgotten a component (such as "self") in the list of
                usable components.
                ... snipped
                It looks like MPI_INIT failed for some reason; your
parallel process is
                likely to abort. There are many reasons that a parallel
process can
                fail during MPI_INIT; some of which are due to
configuration or
                environment
                problems. This failure appears to be an internal
failure; here's some
                additional information (which may only be relevant to an
Open MPI
                developer):
                
                  PML add procs failed
                  --> Returned "Unreachable" (-12) instead of "Success"
(0)
        
------------------------------------------------------------------------
                --
                *** An error occurred in MPI_Init
                *** before MPI was initialized
                *** MPI_ERRORS_ARE_FATAL (goodbye)
                mpirun noticed that job rank 0 with PID 10071 on node
node001 exited on
                signal 1 (Hangup).
                
                
                I would not expect mx messages as communication should
not go through
                the mx card? (This is a twin dual core shared memory
node)
                The same happens when testing on 2 nodes, using a
hostfile.
                I checked the state of the mx card with mx_endpoint_info
and mx_info,
                they are healthy and free.
                What is missing here?
                
                Thanks
                
                Henk
                
                _______________________________________________
                users mailing list
                users_at_[hidden]
                http://www.open-mpi.org/mailman/listinfo.cgi/users