[gmx-users] trjconv gets stuck on frame

Mark Abraham mark.j.abraham at gmail.com
Mon Oct 27 14:07:58 CET 2014


Hi,

Parallel file systems do give uneven service (e.g. one file has a chunk
that lives somewhere that was under high load right when you asked for
it...), so given that you can read the files normally on a normal
filesystem, then you should double-check Stampede's user guides for how to
make best use of their file systems, and/or take up the issue further with
the admins there.

Cheers,

Mark

On Mon, Oct 27, 2014 at 11:40 AM, Eric Smoll <ericsmoll at gmail.com> wrote:

> Hi Mark,
>
> i understand. It wasn't getting stuck in one place, if I skip over the
> problem time when executed from the beginning the slowdown still occurs.
>
> I am working on the Stampede Supercomputer using their install of gromacs
> and for reasons I do not understand, this extremely slow processing only
> occurs for some trajectories on Stampede. Thinking there was some problem
> with the login nodes, I tried this on a compute node with the same results
> - slow processing.
>
> I transferred my trajectory to my laptop, installed the same version of
> gromacs, and processed it with trjconv at a reasonable speed.
>
> Best,
> Eric
>
>
> On Mon, Oct 27, 2014 at 5:33 AM, Mark Abraham <mark.j.abraham at gmail.com>
> wrote:
>
> > On Sun, Oct 26, 2014 at 11:44 PM, Eric Smoll <ericsmoll at gmail.com>
> wrote:
> >
> > > Hi Mark,
> > >
> > > Thank you for responding so rapidly. I should note that identical
> > > processing (I use a script) on the trajectories produced by slightly
> > > different chemical systems had no problem and trajconv produced a
> > complete
> > > processed trajectory.
> > >
> > > However, when processing the problematic few with trajconv, the
> > trajectory
> > > that is output is incomplete (the trjconv output has fewer frames than
> > the
> > > input trajectory).
> > >
> > > This is definitely not problem with the change in output frequency of
> > > progress reports to the terminal.
> > >
> > > I am not sure if the -b flag is telling me anything. I move it around
> and
> > > it still seems to get stuck. I have ~30,000 atoms in my system. The
> first
> > > 120 ps are processed in ~ 5 seconds. The next 4 ps take ~ 30 sec. My
> > > trajectory is many nanoseconds long.
> > >
> >
> > The point is to try to see whether the issue happens x steps into the
> > trajectory, or only at around t=120ps. Does it happen if you are using
> -pbc
> > somethingelse? Does it happen if you copy the file to some other
> filesystem
> > before using -pbc whole? One needs to find a pattern before one can guess
> > where the problem might lie.
> >
> >
> > > Again, my other chemically similar systems do no hang like this and the
> > > simulation procedure is scripted so it is consistent across my
> different
> > > chemical systems.
> > >
> >
> > OK. It's possible your simulation system is doing something pathological
> in
> > that trajectory, which somehow does not agree with the implementation of
> > -pbc whole (I'm guessing wildly here), but one would need to try the
> above
> > kinds of experiments to probe that, and or visualize the trajectory in
> some
> > viewing program.
> >
> > Mark
> >
> > I am using gromacs/4.6.5.
> > >
> > > Best,
> > > Eric
> > >
> > > On Sun, Oct 26, 2014 at 5:21 PM, Mark Abraham <
> mark.j.abraham at gmail.com>
> > > wrote:
> > >
> > > > Hi,
> > > >
> > > > The output does drop in frequency at some point, so that might be all
> > you
> > > > are seeing. Experiment with -b and values around the putative problem
> > > area.
> > > >
> > > > Mark
> > > > On Oct 26, 2014 6:59 PM, "Eric Smoll" <ericsmoll at gmail.com> wrote:
> > > >
> > > > > Hello Gromacs users,
> > > > >
> > > > > I have a trajectory file script18_o.trr that I am trying to
> process.
> > > > Using
> > > > > gmxcheck, this file appears to be complete. When I execute the
> > command
> > > > > below
> > > > >
> > > > > trjconv -f ../script18/script18_o.trr -s ../script17/script17_o.tpr
> > -o
> > > > > tmp1.trr -pbc whole << EOF
> > > > > 0
> > > > > EOF
> > > > >
> > > > > the code moves quickly through the first few hundred frames only to
> > > > > consistently get stuck on frame 300...
> > > > >
> > > > > trn version: GMX_trn_file (single precision)
> > > > >  ->  frame    320 time  128.000        ->  frame    300 time
> 120.000
> > > > >
> > > > > How do I troubleshoot the problem?
> > > > >
> > > > > -Eric
> > > > > --
> > > > > 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.
> > > > >
> > > > --
> > > > 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.
> > > >
> > > --
> > > 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.
> > >
> > --
> > 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.
> >
> --
> 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.
>


More information about the gromacs.org_gmx-users mailing list