[mvapich-discuss] Occasional failure initializing

[Hari Subramoni]

Hello Martin,

Could you please tell us what launcher you're using?

Thx,
Hari.
On Jul 27, 2015 6:13 PM, "Martin Pokorny" <mpokorny at nrao.edu> wrote:

> I'm currently in the process of upgrading an mvapich2 installation from
> version 1.9a2 to version 2.1. Mostly successful, so far, but there is one
> odd issue that I've encountered. I can work around the following issue by
> setting MV2_USE_SHMEM_COLL=0, but I was hoping not to have to do that on a
> continuing basis.
>
> The attached program is sufficient to trigger the problem -- you'll notice
> that it's trivial. Also attached are a host file, a config file, and a
> backtrace. From the backtrace you can see that the failure occurs in the
> call to MPI_Init_thread. I have a core file that I can send, in case that's
> interesting. Note that I've only seen the problem when running in MPMD mode
> using a config file, which matches my case of interest, but I'm not sure
> that's strictly necessary. In this test case, I'm simply providing the same
> executable under two names. Also note that the problem only occurs in about
> one out of ten or twenty trials. Various other settings can change the
> frequency of occurrence, but I figure that's just a further of sign of the
> non-deterministic nature of the problem.
>
> And here's the output of "mpiname -a":
>
>  $ mpiname -a
>> MVAPICH2 2.1 Fri Apr 03 20:00:00 EDT 2015 ch3:mrail
>>
>> Compilation
>> CC: gcc    -DNDEBUG -DNVALGRIND -O2
>> CXX: g++   -DNDEBUG -DNVALGRIND -O2
>> F77: gfortran -L/lib -L/lib   -O2
>> FC: gfortran   -O2
>>
>> Configuration
>> --prefix=/opt/cbe-local/stow/mvapich2-2.1 --enable-romio
>> --with-file-system=lustre --with-limic2
>>
>
> --
> Martin
>
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20150727/27baf61e/attachment-0001.html>