Open MPI logo

Open MPI User's Mailing List Archives

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

Subject: Re: [OMPI users] A problem with 'mpiexec -launch-agent'
From: Reuti (reuti_at_[hidden])
Date: 2010-06-14 18:13:10


Hi,

Am 13.06.2010 um 09:02 schrieb Zhang Linbo:

> Hi,
>
> I'm new to OpenMPI and have encountered a problem with mpiexec.
>
> Since I need to set up the execution environment for OpenMPI programs
> on the execution nodes, I use the following command line to launch an
> OMPI program:
>
> mpiexec -launch-agent /some_path/myscript ....
>
> The problem is: the above command works fine if I invoke 'mpiexec'
> without an absolute path just like above (assuming the PATH variable
> is properly set), but if I prepend an absolute path to 'mpiexec',
> e.g.:
>
> /OMPI_dir/bin/mpiexec -launch-agent /some_path/myscript ....

using an absolute path is equivalent to use the --prefix option to
`mpiexec`. Both ways lead obviously to the erroneous behavior you
encounter.

> then I get the following error message:
>
> bash: -c: line 0: syntax error near unexpected token `('
> bash: -c: line 0: ` PATH=/OMPI_dir/bin:$PATH ; export PATH ;
> LD_LIBRARY_PATH=/OMPI_dir/lib:$LD_LIBRARY_PATH ; export
> LD_LIBRARY_PATH ; /some_path/myscript /OMPI_dir/bin/(null) --
> daemonize -mca ess env -mca orte_ess_jobid 1978662912 -mca
> orte_ess_vpid 1 -mca orte_ess_num_procs 2 --hnp-uri
> "1978662912.0;tcp://180.0.14.12:54844;tcp://190.0.14.12:54844"'

Reason seems to be, that in case of a given prefix the assembly of the
necessary command line includes some elements too much. I tried to
circumvent this by a new case in "orte/mca/plm/rsh/plm_rsh_module.c":

             if (orted_prefix != NULL) {
             asprintf (&final_cmd,
                       "%s%s%s PATH=%s/%s:$PATH ; export PATH ; "
                       "LD_LIBRARY_PATH=%s/%s:$LD_LIBRARY_PATH ;
export LD_LIBRARY_PATH ; "
                       "%s",
                       (opal_prefix != NULL ? "OPAL_PREFIX=" : ""),
                       (opal_prefix != NULL ? opal_prefix : ""),
                       (opal_prefix != NULL ? " ; export
OPAL_PREFIX;" : ""),
                       prefix_dir, bin_base,
                       prefix_dir, lib_base,
                       orted_prefix );
             }
             else {
             asprintf (&final_cmd,
                       "%s%s%s PATH=%s/%s:$PATH ; export PATH ; "
                       "LD_LIBRARY_PATH=%s/%s:$LD_LIBRARY_PATH ;
export LD_LIBRARY_PATH ; "
                       "%s %s/%s/%s",
                       (opal_prefix != NULL ? "OPAL_PREFIX=" : ""),
                       (opal_prefix != NULL ? opal_prefix : ""),
                       (opal_prefix != NULL ? " ; export
OPAL_PREFIX;" : ""),
                       prefix_dir, bin_base,
                       prefix_dir, lib_base,
                       (orted_prefix != NULL ? orted_prefix : ""),
                       prefix_dir, bin_base,
                       orted_cmd);
             }

The name of the agent is for sake of easiness stored in "opal_prefix"
AFAICS.

This is of course not a clean solution (as "opal_prefix" can't be used
any more), but more a proof of concept, as only sh-like shelle are
handled. Sure there are better ways to solve it. Anyway, it's a bug
and should be filed

-- Reuti

> I'd like to know what causes the above problem and how should I deal
> with it.
> I want to use absolute pathname of mpiexec to avoid possible
> inteferences
> with other MPI installations. Thanks in advance.
>
> LB
>
>
> _______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users