"mpirun -gdb" produces a time out

[Deji Akingunola <dakingun@…>]

---------- Forwarded message ----------
From:  <bugzilla@redhat.com>
Date: Sun, Jul 5, 2009 at 7:34 PM
Subject: [Bug 509764] New: "mpirun -gdb" produces a time out
To: dakingun@gmail.com


Please do not reply directly to this email. All additional
comments should be made in the comments box of this bug.

Summary: "mpirun -gdb" produces a time out

https://bugzilla.redhat.com/show_bug.cgi?id=509764

          Summary: "mpirun -gdb" produces a time out
          Product: Fedora
          Version: 11
         Platform: All
       OS/Version: Linux
           Status: NEW
         Severity: high
         Priority: low
        Component: mpich2
       AssignedTo: dakingun@gmail.com
       ReportedBy: laurent.aguerreche@irisa.fr
        QAContact: extras-qa@fedoraproject.org
               CC: dakingun@gmail.com
  Estimated Hours: 0.0
   Classification: Fedora


Description of problem:
The command "mpirun -gdb" does not work and return these messages:
$ mpirun -gdb my-application
0: /usr/bin/mpdgdbdrv.py:20: DeprecationWarning: The popen2 module is
deprecated.  Use the subprocess module.
0:   from popen2 import Popen4
0:  mpdgdbdrv (<module> 107): timed out waiting for initial Breakpoint response

So, it is impossible to debug anything.


How reproducible:
Always.


Any idea?

--
Configure bugmail: https://bugzilla.redhat.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
You are the assignee for the bug.

[Deji Akingunola]

This message has 0 attachment(s)

[goodell]

Hi Deji,

What version of mpich2 has this problem? I suspect that the change to reduce MPI_Init times with mpd that is on the trunk but not in 1.1.0 might be causing this. If you are using 1.1.0 then I don't have any ideas off the top of my head but I will look into this either way.

-Dave

[Deji Akingunola]

Hi Dave,

This bug was originally reported against the MPICH2 package built for
Fedora 11, which implies it cannot be higher than ver. 1.1.0 (I've
asked the original reporter to specify the exact version).

Although running 'mpirun -gdb' works fine for me on Fedora 11, I've
run into similar issue reported here on Centos-5.2 based cluser, with
MPICH2-1.1.0 that was built with the same configuration as for the
binaries in Fedora 11 and running the same application. The only
difference is that I was trying to run the application across multiple
hosts on Centos-5.2, but on just a single single on Fedora 11.

Deji

On Wed, Jul 8, 2009 at 3:02 PM, mpich2<mpich2-maint@mcs.anl.gov> wrote:
> --------------------------------------------------+-------------------------
>  Reporter:  Deji Akingunola <dakingun@gmail.com>  |        Owner:  goodell
>     Type:  bug                                   |       Status:  assigned
>  Priority:  major                                 |    Milestone:
 mpich2-1.1.1
> Component:  mpich2                                |   Resolution:
>  Keywords:                                        |
> --------------------------------------------------+-------------------------
>
> Changes (by goodell):
>
>  * owner:  => goodell
>  * status:  new => assigned
>  * milestone:  => mpich2-1.1.1
>
> Comment:
>
>  Hi Deji,
>
>  What version of mpich2 has this problem?  I suspect that the change to
>  reduce MPI_Init times with mpd that is on the trunk but not in 1.1.0 might
>  be causing this.  If you are using 1.1.0 then I don't have any ideas off
>  the top of my head but I will look into this either way.
>
>  -Dave
>
> --
> Ticket URL: <https://trac.mcs.anl.gov/projects/mpich2/ticket/697#comment:2>
>

[Deji Akingunola]

Hi

>
> Although running 'mpirun -gdb' works fine for me on Fedora 11, I've
> run into similar issue reported here on Centos-5.2 based cluser, with
> MPICH2-1.1.0 that was built with the same configuration as for the
> binaries in Fedora 11 and running the same application. The only

I've taken a second look at my problem with mpdgdbdrv on Centos-5.2,
and it's not the same as being reported here.
The exact error message is as below, let me know if you want me to
open a new ticket about it.

...
0: Traceback (most recent call last):
0:   File "/state/partition1/local/mpich2/bin/mpdgdbdrv.py", line 75, in ?
0:     write(gdb_sin_fileno,'set confirm off\n')
0: OSError: [Errno 32] Broken pipe
maingemclimdm_Linux_x86-64_3.3.0.Abs
...

Deji

> difference is that I was trying to run the application across multiple
> hosts on Centos-5.2, but on just a single single on Fedora 11.
>
> Deji
>
> On Wed, Jul 8, 2009 at 3:02 PM, mpich2<mpich2-maint@mcs.anl.gov> wrote:
>> --------------------------------------------------+-------------------------
>>  Reporter:  Deji Akingunola <dakingun@gmail.com>  |        Owner:  goodell
>>     Type:  bug                                   |       Status:  assigned
>>  Priority:  major                                 |    Milestone:
 mpich2-1.1.1
>> Component:  mpich2                                |   Resolution:
>>  Keywords:                                        |
>> --------------------------------------------------+-------------------------
>>
>> Changes (by goodell):
>>
>>  * owner:  => goodell
>>  * status:  new => assigned
>>  * milestone:  => mpich2-1.1.1
>>
>> Comment:
>>
>>  Hi Deji,
>>
>>  What version of mpich2 has this problem?  I suspect that the change to
>>  reduce MPI_Init times with mpd that is on the trunk but not in 1.1.0 might
>>  be causing this.  If you are using 1.1.0 then I don't have any ideas off
>>  the top of my head but I will look into this either way.
>>
>>  -Dave
>>
>> --
>> Ticket URL: <https://trac.mcs.anl.gov/projects/mpich2/ticket/697#comment:2>
>>
>

[goodell]

These deprecation warning messages were suppressed in r4923. Darius is looking into the centos "Broken Pipe" error.

-Dave

[balaji]

Darius wasn't aware that this was on his plate. Are we still planning this for 1.1.1?

[goodell]

I don't think it's essential for 1.1.1. I'm fine with fixing this for 1.1.2 instead.

[buntinas]

This bug has been fixed in r4979. I'll open a new ticket for the centos bug.