[gmx-users] Problem mdrun_mpi GROMACS 4.6.1 in Intel Cluster

[Mark Abraham]

On Thu, Nov 20, 2014 at 2:32 AM, Agnivo Gosai <agnivogromacs14 at gmail.com>
wrote:

> Dear Users
>
> I am again back with my issue of GROMACS 4.6.X and the Intel cluster used
> for my research.
>
> This time ,I compiled a serial ( without any MPI threading version ) of
> GROMACS 4.6.7 by the following command line :-
>
> /work/gb_lab/agosai/GROMACS/cmake-2.8.11/bin/cmake .. -DGMX_GPU=OFF
> -DGMX_MPI=OFF -DGMX_OPENMP=OFF -DGMX_THREAD_MPI=OFF -DGMX_OPENMM=OFF
> -DCMAKE_C_COMPILER=icc -DCMAKE_CXX_COMPILER=icpc -DGMX_BUILD_OWN_FFTW=ON
> -DCMAKE_INSTALL_PREFIX=/work/gb_lab/agosai/gmx467ag -DGMX_DOUBLE=ON
>
> Now , I am not getting any problems with "genbox" , "grompp" and other
> GROMACS tools. I hope I will not face any errors like the one mentioned in
> the *trailing mail*.
>
> For mdrun I use the mdrun_mpi installed by a colleague of mine using
> GROMACS 4.6.1 package. I am using serial commands from 4.6.7 for pre and
> post processing and mdrun_mpi from 4.6.1 for the molecular dynamics runs.
>
>  I observe that my cluster has got 1 node having 16 processors. Using
> mpirun -np 16 mdrun_mpi ..... gives desired result with on an average 6.5
> ns / day of simulation time.
>
> However when I use more than 1 node (4*16 = np = 64) , there are no output
> files generated by mdrun_mpi. Nothing happens. I deleted the job and found
> the following errors.
> [mpiexec at node094] HYD_pmcd_pmiserv_send_signal
> (./pm/pmiserv/pmiserv_cb.c:221): assert (!closed) failed
> [mpiexec at node094] ui_cmd_cb (./pm/pmiserv/pmiserv_pmci.c:128): unable to
> send SIGUSR1 downstream
> [mpiexec at node094] HYDT_dmxu_poll_wait_for_event
> (./tools/demux/demux_poll.c:77): callback returned error status
> [mpiexec at node094] HYD_pmci_wait_for_completion
> (./pm/pmiserv/pmiserv_pmci.c:388): error waiting for event
> [mpiexec at node094] main (./ui/mpich/mpiexec.c:718): process manager error
> waiting for completion
> ~
>
> I believe that mdrun_mpi from version 4.6.1 has not been properly compiled
> to run on several nodes in the cluster. However I would like to do it to
> speed up my simulations.
>
> Any suggestions , please ??
>

You're going to need to read your local cluster documentation and speak to
your system adminstrators for the right information to set up your MPI
compilation and/or environment, like Justin has already suggested to you.
Every site is different, so you can't usefully ask people on the other side
of the world... And please do it with 4.6.7, for your own good ;-)

Mark


>
> Thanks & Regards
> Agnivo Gosai
> Grad Student, Iowa State University.
>
> ---------- Forwarded message ----------
> From: Mark Abraham <mark.j.abraham at gmail.com>
> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> Cc:
> Date: Mon, 10 Nov 2014 02:26:09 +0100
> Subject: Re: [gmx-users] New problem GROMACS 4.6.7 in Intel Cluster
> On Mon, Nov 10, 2014 at 2:17 AM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
> >
> >
> > On 11/9/14 8:15 PM, Agnivo Gosai wrote:
> >
> >> Dear Users
> >>
> >> Firstly thanks (specially Drs. Szilard and Mark ) for helping me out
> with
> >> the installation of GROMACS 4.6.7 on my university Intel cluster.
> However
> >> I
> >> ran into a new problem while using it.
> >>
> >> 1. Firstly I used pdb2gmx to process a pdb file.
> >> 2. Then I used editconf.
> >> 3. Then I used genbox.
> >>
> >> But I encountered with the following error.
> >>
> >> Initialising van der waals distances...
> >> Will generate new solvent configuration of 5x5x9 boxes
> >> Generating configuration
> >> Sorting configuration
> >> Found 1 molecule type:
> >>      SOL (   3 atoms): 48600 residues
> >> Calculating Overlap...
> >> box_margin = 0.315
> >> Removed 53454 atoms that were outside the box
> >> OMP: Error #178: Function pthread_getattr_np failed:
> >> OMP: System error #12: Cannot allocate memory
> >> Aborted
> >>
> >> Now upon searching on the web it seems to be an OpenMP error. Again I am
> >> in
> >> a fix which I have little or no idea about.
> >>
> >>
> > Preprocessing tools (and most analysis tools) are not parallelized in any
> > way. They run on one core only.  Generally you do not carry out these
> sorts
> > of operations on a cluster, as there is no benefit to doing so
> >
> > -Justin
>
> True, and the particular problem is likely that Agnivo has not prepared the
> environment correctly, e.g. by sourcing Intel's compilervars.sh script,
> and/or following the cluster's usage guide for loading its Intel module.
>
> Mark
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>