Hi Jason,
Thanks for the reply. As it turns out it is because in the PBS script I
used the files were copied to a scratch directory on the master-node. This
caused the slave nodes not to have access to the files and thus they sat
idle... I've fixed it by running directory from my HOME directory, which
all the nodes have access to.
Best regards,
Jesper
On 12/01/11 19.23, "Jason Swails" <jason.swails.gmail.com> wrote:
>With the same setup, we've been able to run just fine across different
>nodes.
>
>Look at the options that exist for mpirun -- I'm guessing it's not getting
>the proper machinefile or nodefile (I'm not an expert in this), so it
>doesn't know to start threads on multiple machines.
>
>If your cluster has PBS, then that often takes care of it for you, but
>possibly not. Try reading through some of the OpenMPI documentation to
>see
>if it gives you any leads (or see your system administrator for some
>advice). It may also be worth testing to see if you can reproduce this
>behavior with other MPI programs compiled the same way.
>
>Good luck!
>Jason
>
>2011/1/12 Jesper Sørensen <lists.jsx.dk>
>
>> Hello,
>>
>>
>>
>> I have installed MMPBSA.py on our cluster and the parallel versions
>>works
>> well on a single node using all 8 CPU¹s.
>>
>> However, once I perform a calculation using several nodes, it only
>>starts
>> on
>> the master-node leaving the other nodes running nothing.
>>
>>
>>
>> I am using the following versions of programs
>>
>> Amber 11 with all patches
>>
>> AmberTools 1.4 with all patches
>>
>> Mpi4py-1.2.2
>>
>> ActivePython 2.7.1.3
>>
>> OpenMPI 1.4.1
>>
>> mmpbsa_py (downloaded from the tutorial website) with all patches
>>
>>
>>
>> I have tried running the calculation on nodes using either Ethernet or
>> infiniband, but the error occurs using both setups.
>>
>>
>>
>> All suggestions are appreciated.
>>
>>
>>
>> Best regards,
>>
>>
>>
>> Jesper Soerensen
>>
>> Ph.D. Student
>>
>> Department of Chemistry
>>
>> Aarhus University, Denmark
>>
>> jes.chem.au.dk
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
>
>
>--
>Jason M. Swails
>Quantum Theory Project,
>University of Florida
>Ph.D. Graduate Student
>352-392-4032
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Jan 16 2011 - 11:00:02 PST
