[mvapich-discuss] mpirun_rsh issue with different users on the same computers

[Krishna Chaitanya Kandalla]

Robert,
          Good to know that you are now able to run your jobs with this 
flag turned off. Since both the jobs contend for the same cores, its 
quite possible that they were making very slow progress.

Thanks,
Krishna

Robert Soliday wrote:
> Thank you. By adding MV2_ENABLE_AFFINITY=0 to the command it solved 
> the problem. I wouldn't have expected this because I didn't think the 
> jobs were making any progress before but perhaps I was wrong.
>
> --Bob Soliday
>
> Krishna Chaitanya Kandalla wrote:
>> Robert,             Once you have started the second job, are the two 
>> jobs making any progress? Can you try disabling cpu affinity by 
>> setting MV2_ENABLE_AFFINITY=0 while launching both the jobs. In 
>> MVAPICH2, cpu_affinity is on by default and we deal with mapping 
>> processes to cores. However, if you are running more than one job on 
>> the same compute node, its very likely that the processes will get 
>> scheduled on the same cores. In such cases, we are better off letting 
>> the linux kernel take care of mapping the processes to cores. 
>> Ofcourse, this is under the assumption that you are having a few 
>> "idle" cores in each node when you start the second job.  Also, which 
>> version of MVAPICH/MVAPICH2 are you using?
>>
>> Thanks,
>> Krishna
>>
>> Robert Soliday wrote:
>>> If I launch a job with mpirun_rsh it runs without issues. But if 
>>> while I am running this job, another user starts another job with 
>>> mpirun_rsh and it lands on some of the same nodes then both jobs 
>>> pause until one of them is killed. Then the other that was not 
>>> killed finishes normally. Is there a port conflict or something that 
>>> I should be looking for?
>>>
>>> Thanks,
>>> --Bob
>>> _______________________________________________
>>> mvapich-discuss mailing list
>>> mvapich-discuss at cse.ohio-state.edu
>>> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>>>
>>>
>
>