[gmx-users] query for gromacs-4.5.4

Mark Abraham mark.j.abraham at gmail.com
Thu Mar 14 14:20:58 CET 2013 

One cannot run an arbitrary simulation on an arbitrary number of
processors. The particles in the simulation have to get mapped to
processors that handle them. This is done in small parcels called domains.
The atoms move between domains, and so the responsibility moves too. If the
velocities are high with respect to the size of the domains, the
responsibility might shift too far for the implementation to handle. This
is happening here. GROMACS does not handle arbitrary redistribution of
particles to processors because the algorithm to do that would be too
expensive for the normal case where the redistribution is more orderly.

So either the velocities are too high ("the system is not well
equilibrated"), and you are getting lucky on small numbers of nodes that
the domains are big enough to cope with the equilibration process, or you
are trying to use domains that are just too small for normal velocities.

We'd need to know what's in the .tpr to have an opinion about what a
reasonable maximum number of nodes might be, but a simulation with less
than a thousand atoms per processor  starts having to pay attention to
these issues.

Mark

On Thu, Mar 14, 2013 at 1:25 PM, Chaitali Chandratre
<chaitujoshi at gmail.com>wrote:

> Hello Sir,
>
> The job runs for 8 processes given 1 , 2 or 8 nodes but not for more than
> that.
> 16 proceses : Segmentation fault and
> For 32 processes it gives :
> "Fatal error : 467 particles communicated to PME node 4 are more than 2/3
> times
> the cut off out of the domain decomposition cell of their charge group in
> deimesion x.
> This usually means that your system is not well equilibrated"
>
> What can be the reason...
>
> Thanks,
> Chaitali
>
> On Tue, Mar 12, 2013 at 3:50 PM, Mark Abraham <mark.j.abraham at gmail.com
> >wrote:
>
> > It could be anything. But until we see some GROMACS diagnostic messages,
> > nobody can tell.
> >
> > Mark
> >
> > On Tue, Mar 12, 2013 at 10:08 AM, Chaitali Chandratre <
> > chaitujoshi at gmail.com
> > > wrote:
> >
> > > Sir,
> > >
> > > Thanks for your reply....
> > > But the same script runs on some other cluster with apprx same
> > > configuration but not on cluster on which I am doing set up.
> > >
> > > Also job hangs after some 16000 steps but not come out immediately.
> > > It might be problem with configuration or what?
> > >
> > > Thanks...
> > >
> > > Chaitali
> > >
> > > On Tue, Mar 12, 2013 at 2:18 PM, Mark Abraham <
> mark.j.abraham at gmail.com
> > > >wrote:
> > >
> > > > They're just MPI error messages and don't provide any useful GROMACS
> > > > diagnostics. Look in the end of the .log file, stderr and stdout for
> > > clues.
> > > >
> > > > One possibility is that your user's system is too small to scale
> > > > effectively. Below about 1000 atoms/core you're wasting your time
> > unless
> > > > you've balanced the load really well. There is a
> > > > simulation-system-dependent point below which fatal GROMACS errors
> are
> > > > assured.
> > > >
> > > > Mark
> > > >
> > > > On Tue, Mar 12, 2013 at 6:17 AM, Chaitali Chandratre
> > > > <chaitujoshi at gmail.com>wrote:
> > > >
> > > > > Hello Sir,
> > > > >
> > > > > Actually I have been given work to setup gromacs-4.5.4 in our
> cluster
> > > > which
> > > > > is being used
> > > > > by users.I am not gromacs user and not aware of its internal
> details.
> > > > > I have got only .tpr file from user and I need to test my
> > installation
> > > > > using that .tpr file.
> > > > >
> > > > > It works fine for 2 nodes 8 processes , 1 node 8 processes.
> > > > >  But when number of processes are equal to 16 it gives segmentation
> > > fault
> > > > > and
> > > > >  if number of processes are equal to 32 it gives
> > > > > error message like
> > > > > " HYD_pmcd_pmiserv_send_signal (./pm/pmiserv/pmiserv_cb.c:221):
> > assert
> > > > > (!closed) failed
> > > > >  ui_cmd_cb (./pm/pmiserv/pmiserv_pmci.c:128): unable to send
> SIGUSR1
> > > > > downstream
> > > > >  HYDT_dmxu_poll_wait_for_event (./tools/demux/demux_poll.c:77):
> > > callback
> > > > > returned error status
> > > > >  HYD_pmci_wait_for_completion (./pm/pmiserv/pmiserv_pmci.c:388):
> > error
> > > > > waiting for event
> > > > > [ main (./ui/mpich/mpiexec.c:718): process manager error waiting
> for
> > > > > completion"
> > > > >
> > > > > I am not clear like whether problem is there in my installation or
> > > what?
> > > > >
> > > > > Thanks and Regards,
> > > > >    Chaitalij
> > > > >
> > > > > On Wed, Mar 6, 2013 at 5:41 PM, Justin Lemkul <jalemkul at vt.edu>
> > wrote:
> > > > >
> > > > > >
> > > > > >
> > > > > > On 3/6/13 4:20 AM, Chaitali Chandratre wrote:
> > > > > >
> > > > > >> Dear Sir ,
> > > > > >>
> > > > > >> I am new to this installation and setup area. I need some
> > > information
> > > > > for
> > > > > >> -stepout option for
> > > > > >>
> > > > > >
> > > > > > What more information do you need?
> > > > > >
> > > > > >
> > > > > >  mdrun_mpi and also probable causes for segmentation fault in
> > > > > >>  gromacs-4.5.4.
> > > > > >> (my node having 64 GB mem running with 16 processes, nsteps =
> > > > 20000000)
> > > > > >>
> > > > > >>
> > > > > > There are too many causes to name.  Please consult
> > > > > http://www.gromacs.org/
> > > > > > **Documentation/Terminology/**Blowing_Up<
> > > > > http://www.gromacs.org/Documentation/Terminology/Blowing_Up>.
> > > > > >  If you need further help, please be more specific, including a
> > > > > description
> > > > > > of the system, steps taken to minimize and/or equilibrate it, and
> > any
> > > > > > complete .mdp file(s) that you are using.
> > > > > >
> > > > > > -Justin
> > > > > >
> > > > > > --
> > > > > > ==============================**==========
> > > > > >
> > > > > > Justin A. Lemkul, Ph.D.
> > > > > > Research Scientist
> > > > > > Department of Biochemistry
> > > > > > Virginia Tech
> > > > > > Blacksburg, VA
> > > > > > jalemkul[at]vt.edu | (540) 231-9080
> > > > > > http://www.bevanlab.biochem.**vt.edu/Pages/Personal/justin<
> > > > > http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin>
> > > > > >
> > > > > > ==============================**==========
> > > > > > --
> > > > > > gmx-users mailing list    gmx-users at gromacs.org
> > > > > > http://lists.gromacs.org/**mailman/listinfo/gmx-users<
> > > > > http://lists.gromacs.org/mailman/listinfo/gmx-users>
> > > > > > * Please search the archive at http://www.gromacs.org/**
> > > > > > Support/Mailing_Lists/Search<
> > > > > http://www.gromacs.org/Support/Mailing_Lists/Search>before
> posting!
> > > > > > * Please don't post (un)subscribe requests to the list. Use the
> www
> > > > > > interface or send it to gmx-users-request at gromacs.org.
> > > > > > * Can't post? Read
> http://www.gromacs.org/**Support/Mailing_Lists<
> > > > > http://www.gromacs.org/Support/Mailing_Lists>
> > > > > >
> > > > > --
> > > > > gmx-users mailing list    gmx-users at gromacs.org
> > > > > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > > > * Please search the archive at
> > > > > http://www.gromacs.org/Support/Mailing_Lists/Search before
> posting!
> > > > > * Please don't post (un)subscribe requests to the list. Use the
> > > > > www interface or send it to gmx-users-request at gromacs.org.
> > > > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > > > >
> > > > --
> > > > gmx-users mailing list    gmx-users at gromacs.org
> > > > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > > * Please search the archive at
> > > > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > > > * Please don't post (un)subscribe requests to the list. Use the
> > > > www interface or send it to gmx-users-request at gromacs.org.
> > > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > > >
> > > --
> > > gmx-users mailing list    gmx-users at gromacs.org
> > > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > * Please search the archive at
> > > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > > * Please don't post (un)subscribe requests to the list. Use the
> > > www interface or send it to gmx-users-request at gromacs.org.
> > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > >
> > --
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > * Please don't post (un)subscribe requests to the list. Use the
> > www interface or send it to gmx-users-request at gromacs.org.
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>

More information about the gromacs.org_gmx-users mailing list