[mpich-discuss] mpiexec (hydra) can't run in the background

[Manhui Wang]

Dear Pavan,

I still see the similar problem with latest MPICH2-1.3.2:

 mpicc hello_world_wait.c -o hello.1.3.2
[sacmw4 at arccacluster11 testprogram]$ mpiexec -np 2 ./hello.1.3.2
>output132.log &
[1] 12659
[sacmw4 at arccacluster11 testprogram]$ [mpiexec at arccacluster11]
HYDU_sock_read (./utils/sock/sock.c:222): read errno (Input/output error)
[mpiexec at arccacluster11] control_cb (./pm/pmiserv/pmiserv_cb.c:249):
assert (!closed) failed
[mpiexec at arccacluster11] HYDT_dmxu_poll_wait_for_event
(./tools/demux/demux_poll.c:77): callback returned error status
e[mpiexec at arccacluster11] HYD_pmci_wait_for_completion
(./pm/pmiserv/pmiserv_pmci.c:206): error waiting for event
[mpiexec at arccacluster11] main (./ui/mpich/mpiexec.c:404): process
manager error waiting for completion
cho

[1]+  Exit 255                mpiexec -np 2 ./hello.1.3.2 > output132.log


The bug related to this
http://trac.mcs.anl.gov/projects/mpich2/ticket/1029

has been closed saying "Fixed in r7485 Resolving". I have a copy of
mpich2-trunk-r7524, and it does work. Presumably the fix is not
committed in MPICH2-1.3.2 yet, or later patches caused this problem?

Could you please take a look at it?

Thank you very much.
Manhui


Pavan Balaji wrote:
> 
> Thanks for the report. I can reproduce this problem. I've reopened the
> ticket and will look into it. For the time being, please continue to
> redirect /dev/null to stdin as a workaround.
> 
>  -- Pavan
> 
> On 10/29/2010 09:51 AM, Manhui Wang wrote:
>> Hello,
>>
>> I have seen a problem with mpiexec when running a simple hello_world
>> (see attached hello_world_wait.c) program in the background. The
>> background job stops after typing any command:
>> ~/testprogram>  mpicc hello_world_wait.c -o hello
>> ~/testprogram>  mpiexec -np 2 ./hello>output.log&
>> [1] 19958
>> ~/testprogram>  echo hello
>> hello
>>
>> [1]+  Stopped                 mpiexec -np 2 ./hello>  output.log
>>
>>
>> It works fine if I run it with</dev/null
>>
>> mpiexec -np 2 ./hello>output.log</dev/null&
>>
>>
>> I am using the latest version (MPICH2 1.3), and the hydra is the default
>> process manage. The problem can be reproduce with mpiexec.hydra for
>> mpich2 1.2.1p1.
>>
>> I saw there was a bug report about this
>> http://trac.mcs.anl.gov/projects/mpich2/ticket/1029
>> ,but now it has been closed/fixed.
>>
>> Any help would be appreciated.
>>
>> Thanks,
>> Manhui
>>
>>
>>
>>
>> _______________________________________________
>> mpich-discuss mailing list
>> mpich-discuss at mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> 

-- 
-----------
Manhui  Wang
School of Chemistry, Cardiff University,
Main Building, Park Place,
Cardiff CF10 3AT, UK