[mpich-discuss] HYDRA and kill process

[Torquil Macdonald Sørensen]

Hi Pavan!

I am running Mpich2 r8281, so it probably contains the newest Hydra at that URL 
at the time of writing of your previous post, since the newest Hydra nightly 
tarball at that time was r8281.

But let me if I understood you correctly before I go to the trouble of building 
another mpich2 and/or hydra: You are saying that I should be able to run (with 
four procs on each host)

mpirun -n 8 nice prog.exe

and it should work properly just as would

mpirun -n 8 prog.exe

I.e, there should not be any trouble with communication between processes on 
different hosts even if I include the "nice" command? It should be possible in 
both cases to stop all processes by hitting CRTL-c on the host at which the 
mpirun command was run?

If that's your claim, I can try it again using the latest nightly tarball of 
mpich2 and hydra. At the moment I am lowering processes priorities internally in 
the program, i.e. using setpriority() in my C++ code. So I don't really need the 
"nice" command, but I can test it, if it is supposed to work.

Thanks
Torquil Sørensen

On 27/03/11 17:03, Pavan Balaji wrote:
> Hi Torquil,
>
> I tried many different combinations for this with different numbers of
> nodes, but it always works correctly for me.
>
> Just so we are not chasing some bug that has already gotten fixed, can
> you try the latest nightly snapshot of Hydra?
>
> http://www.mcs.anl.gov/research/projects/mpich2/downloads/tarballs/nightly/hydra/
>
>
> Thanks,
>
> -- Pavan
>
> On 03/26/2011 02:23 AM, Torquil Macdonald Sørensen wrote:
>> On 25/03/11 15:27, Pavan Balaji wrote:
>>> 3. Did you try running one of the MPICH2 example programs
>>> (./examples/cpi)?
>>
>> The 'cpi' example does not have the same problem.
>>
>> I think I've determined the cause. There is something I forgot to
>> mention in my
>> original description. I was running
>>
>> mpirun -n 8 time nice -n 19 prog.exe
>>
>> in an effort to get low priority processes. Removing 'time' and 'nice'
>> fixes the
>> problem, i.e. using
>>
>> mpirun -n 8 prog.exe
>>
>> The 'time' command I can live without. Now I have to figure out how to
>> set a low
>> process priority without using 'nice'.
>>
>> Thanks for all your help!
>>
>> Best regards
>> Torquil Sørensen
>