[mpich-discuss] kill process => terminate job?

Calin Iaru calin at dolphinics.com
Tue Sep 16 12:25:59 CDT 2008


I don't use MPI_Errhandler_create or any other custom handler. The 
applications are single threaded. I will let you know what happens next 
time:
 - daemon is running or not
 - stack trace on daemon
 - stack trace on process


----- Original Message ----- 
From: "Robert Latham" <robl at mcs.anl.gov>
To: <mpich-discuss at mcs.anl.gov>
Sent: Tuesday, September 16, 2008 5:09 PM
Subject: Re: [mpich-discuss] kill process => terminate job?


> On Sun, Sep 14, 2008 at 06:33:23PM +0200, Calin Iaru wrote:
>> When I run programs under Windows or Linux, I can see that an abrupt
>> termination of one process may lead to the whole job being
>> terminated by the process managers. This is not always the case, as
>> sometimes orphaned processes are still running. Has anyone seen this
>> with mpd or smpd on Windows? Could it be that the behaviour in such
>> cases is not strong, meaning that it is up to the process manager to
>> decide what to do next? If so, then what are the criterias for
>> terminating a job?
>
> What is your error handler set to?  It sounds like the default error
> handler, MPI_ERRORS_ARE_FATAL, is kicking in.
>
> http://www.mpi-forum.org/docs/mpi-11-html/node148.html#Node148
>
> ==rob
>
> -- 
> Rob Latham
> Mathematics and Computer Science Division    A215 0178 EA2D B059 8CDF
> Argonne National Lab, IL USA                 B29D F333 664A 4280 315B
>
> 




More information about the mpich-discuss mailing list