[gmx-users] mdrun_mpi issue with CHARMM36 FF

[Anirban]

Hi ALL,

I am trying to simulate a membrane protein system using CHARMM36 FF on
GROAMCS4.5.5 on a parallel cluster running on MPI. The system consists of
arounf 1,17,000 atoms. The job runs fine on 5 nodes (5X12=120 cores) using
mpirun and gives proper output. But whenever I try to submit it on more
than 5 nodes, the job gets killed with the following error:

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

starting mdrun 'Protein'
50000000 steps, 100000.0 ps.

NOTE: Turning on dynamic load balancing

Fatal error in MPI_Sendrecv: Other MPI error
Fatal error in MPI_Sendrecv: Other MPI error
Fatal error in MPI_Sendrecv: Other MPI error

=====================================================================================
=   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
=   EXIT CODE: 256
=   CLEANING UP REMAINING PROCESSES
=   YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
=====================================================================================
[proxy:0:0 at cn034] HYD_pmcd_pmip_control_cmd_cb
(./pm/pmiserv/pmip_cb.c:906): assert (!closed) failed
[proxy:0:0 at cn034] HYDT_dmxu_poll_wait_for_event
(./tools/demux/demux_poll.c:77): callback returned error status
[proxy:0:0 at cn034] main (./pm/pmiserv/pmip.c:214): demux engine error
waiting for event
.
.
.
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Why is this happening? Is it related to DD and PME? How to solve it? Any
suggestion is welcome.
Sorry for re-posting.


Thanks and regards,

Anirban
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120514/67c5d5f5/attachment.html>