[gmx-users] problem with restarting from cpt and _prev.cpt file in REMD simulation

[Tomek Wlodarski]

Dear Gromacs users,

It seems that now my simulation is not able to write any cpt file...
I had run plain mdrun to advance each simulation to the same time point and
then I restarted replica exchange simulation and it was running without
errors but was not writing any cpt files....
I tried to restart this simulation with mdrun -nsteps  and amazingly it
worked and I got new cpt files but next restarts with or without -nsteps
option are not proving any cpt files... any ideas where may be the problem?
Thank you for any suggestions!
Best,

tomek

On Thu, Oct 22, 2015 at 1:04 PM, Tomek Wlodarski <tomek.wlodarski at gmail.com>
wrote:

> Dear Mark,
>
> Thank you for suggestion.
> I run plain mdrun to advance each simulation to the same time point and
> use new cpt to restart replica exchange simulations.
> It started without any error from the proper time point and metadynamics
> was running OK.
> However, when it finished it did not write any new cpt files... so when I
> restarted it again it started from the the same time point like the one
> before....
> Any idea what may cause this problem?
> Thank you.
> Best,
>
> tomek
>
> On Tue, Oct 13, 2015 at 11:43 AM, Mark Abraham <mark.j.abraham at gmail.com>
> wrote:
>
>> Hi,
>>
>> Unfortunately, that should
>> a) never happen,
>> b) given a), rarely happen identically for successive checkpoints,
>> c) is even more strange if your time step is a normal ~2fs, and
>> d) is not easy to fix.
>>
>> To address a-c), it would be good for you to get some feedback from the
>> PLUMED developers about whether others have reported such problems.
>> Offhand, I don't know of any plain GROMACS user who's been able to produce
>> such checkpoint files. Last time I looked at their code I didn't see
>> anything that could do this, but that was 1.x. It is also conceivable that
>> this is a bug in your MPI layer.
>>
>> For d), you can try
>> * back up your files
>> * use plain mdrun to advance all of the 194440.010 files by one step (e.g.
>> gmx mdrun -nsteps 1 -s old -cpi state)
>> * same for the 194440.002 files by as many steps as you need to catch up
>> to
>> the others
>> * restart mdrun -multi -replex
>>
>> Simulations have an internal part number that is expected to match, which
>> is why you can't just run an extra part to catch up the simulations that
>> are lagging.
>>
>> Mark
>>
>> On Tue, Oct 13, 2015 at 12:05 PM Tomek Wlodarski <
>> tomek.wlodarski at gmail.com>
>> wrote:
>>
>> > Dear GROMACS users,
>> >
>> > I am running REMD simulation with Gromacs 5.0.4 and PLUMED 2.1 for
>> already
>> > around 200ns with many smooth restarts. Unfortunately now something went
>> > wrong because I got error that:
>> >
>> > The 6 subsystems are not compatible
>> >
>> > So I checked with gmxcheck cpt files
>> > replica 1: Last frame         -1 time 194440.010
>> > replica 2: Last frame         -1 time 194440.010
>> > replica 3: Last frame         -1 time 194440.010
>> > replica 4: Last frame         -1 time *194440.002*
>> > replica 5: Last frame         -1 time 194440.010
>> > replica 6: Last frame         -1 time *194440.002*
>> > Same situation is in _prev.cpt
>> > replica 1: Last frame         -1 time 194400.010
>> > replica 2: Last frame         -1 time 194400.010
>> > replica 3: Last frame         -1 time 194400.010
>> > replica 4: Last frame         -1 time *194400.002*
>> > replica 5: Last frame         -1 time 194400.010
>> > replica 6: Last frame         -1 time *194400.002*
>> >
>> > Any idea how to restart simulation when both cpt and _prev.cpt file
>> seem to
>> > be broken?
>> > Why this happened and how to prevent it?
>> > Thank you.
>> > All the best,
>> >
>> > tomek
>> > --
>> > 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.
>>
>
>