[mpich-discuss] Hydra SIGTSTP (Ctrl+Z) handling

[Yauheni Zelenko]

Hi, Pavan!

I checked nightly with modified cpi example and signals (SIGTSTP) are propagated correctly. However cpi was not suspended when running in MPI mode.

Looks like SIGTSTP handling is more complicated then just relaying it to MPI processes...

Eugene.
________________________________________
From: mpich-discuss-bounces at mcs.anl.gov [mpich-discuss-bounces at mcs.anl.gov] On Behalf Of Pavan Balaji [balaji at mcs.anl.gov]
Sent: Wednesday, December 01, 2010 12:51 AM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Hydra SIGTSTP (Ctrl+Z) handling

I've committed an initial attempt at this into trunk in r7530. Can you
try out the nightly snapshot to see if it works for you?

I'm not totally confident of this, as I've disabled the propagation of
all signals to the forked processes (is there a way to disable the
propagation only for select signals?), so all signals have to be
explicitly passed. This means that for signals that are not explicitly
handled by mpiexec, it might terminate, leaving the launcher processes
dangling around.

  -- Pavan

On 11/30/2010 09:50 PM, Pavan Balaji wrote:
>
> SIGTSTP, not SIGINT. Sorry.
>
> On 11/30/2010 09:48 PM, Pavan Balaji wrote:
>>
>> On 11/30/2010 08:00 PM, Yauheni Zelenko wrote:
>>> Why signals could not be trapped in Hydra without relaying to
>>> launchers (child processes if I'm correct) in default way? Then it's
>>> only matter of transferring signals to proxies via socket.
>>
>> Passing the signals to the proxy is not a problem. Not relaying it to
>> the launchers is the problematic part -- maybe I'm missing something
>> straightforward here, but I can't seem to find a way to do this.
>>
>> Hydra's mpiexec forks a process, which internally execvp's SSH. MPIEXEC
>> catches SIGINT and deals with it. However, SSH also catches SIGINT and
>> closes all its sockets. I want to avoid the latter part. Is there a good
>> way to do this?
>>
>>     -- Pavan
>>
>

--
Pavan Balaji
http://www.mcs.anl.gov/~balaji
_______________________________________________
mpich-discuss mailing list
mpich-discuss at mcs.anl.gov
https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss