[gmx-users] Restart MD with overwritten CPT using TRR file in gromacs/2016.3

[Mark Abraham]

Hi,


On Mon, Oct 30, 2017 at 6:55 PM Abramyan, Tigran <tigran at email.unc.edu>
wrote:

> Thanks a lot for your response, Mark,
>
>
> Your response did not come to my email, but I found it via Gromacs
> website. I think my email is registered on your site, but I forgot my
> Username. Is there a way to login with the email address and password? At
> the moment it looks like the only way I can reply to your response is via
> this email? Please suggest the correct way if I am wrong here.
>

Yes, there's information in the footer of every email from the list.


>
> So, as for the issue, it looks like Gromacs has two CPT files, both of
> which got rewritten when the job got resubmitted after crashing. The names
> of these CPT files are md.cpt and md_prev.cpt, and both have the same
> writing time (the current time, as the job is running now). I do not see
> any #md.cpt or #md_prev.cpt files, which I could use to restart the job
> from the time (frame and velocities) when it crashed.
>

Yes, mdrun doesn't back those up, as they tend to be large and of limited
value to retain.

I am thinking of ways to go around this issue and would like to (1) extract
> the last frame from the TRR file and (2) use the TRR file to read the
> velocities from. So, for the first step I do the following:
>
>
> cp #md.trr.1# md.trr  (is this correct, by the way?)
>

Could be, depends if the #md.trr.1# file is sound or not. Use gmx check to
see what it has.

echo "0" | gmx trjconv -f md.trr -s md.tpr -o lasframe.gro -b lastframe -e
> lastframe
>
>
> But I still obtain the same error, which makes me suspect something is
> wrong with the #md.trr.1# file?
>

Yes, anything could have happened before the crash.

Given your experience on this cluster, I recommend that you a) complain to
the sysadmins that the file system is not doing the job well enough, and b)
use mdrun -noappend so that your potential for loss is more limited. You
will now have to concatenate trajectory and energy files yourself, but at
least you will only lose data if you choose to delete it ;-)

Mark


> Program:     gmx trjconv, version 2016.3
> Source file: src/gromacs/fileio/trrio.cpp (line 114)
>
> Fatal error:
> Failed to find GROMACS magic number in trr frame header, so this is not a
> trr
> file!
>
>
>
> --
> Tigran M. Abramyan, Ph.D.
> Postdoctoral Fellow, Computational Biophysics & Molecular Design
> Center for Integrative Chemical Biology and Drug Discovery
> Eshelman School of Pharmacy
> The University of North Carolina at Chapel Hill
> Chapel Hill, NC 27599-7363
>
> ________________________________
> From: Abramyan, Tigran
> Sent: Friday, October 27, 2017 4:24:15 PM
> To: gromacs.org_gmx-users at maillist.sys.kth.se
> Subject: Restart MD with overwritten CPT using TRR file in gromacs/2016.3
>
>
> Hi,
>
>
> Because one of my gromacs/2016.3 jobs accidentally crashed and got
> restarted and overwrote the CPT file, I am now attempting to continue a
> simulation using the TRR file. For this I ran the command 'gmx convert-tpr
> -s md.tpr -f md.trr -o md_next.tpr', so that I could use 'md_next.tpr' file
> to restart the simulation, but obtain the following error:
>
>
> READING COORDS, VELS AND BOX FROM TRAJECTORY md.trr...
>
> trr version: GMX_trn_file (single precision)
> Read    trr frame   7519: step 150380000 time  300760.000
> -------------------------------------------------------
> Program:     gmx convert-tpr, version 2016.3
> Source file: src/gromacs/fileio/trrio.cpp (line 114)
>
> Fatal error:
> Failed to find GROMACS magic number in trr frame header, so this is not a
> trr
> file!
>
>
>
> The same command (only replace 'gmx convert' to older version 'tpbconv')
> works fine in gromacs/4.6.3.
>
>
> Additionally, I was wondering, why doesn't gromacs automatically 'hash'
> backup the CPT file along with the other files? Does that mean that CPT
> file probably has info (e.g., velocities) from the previous run, which
> didn't get overwritten?
>
>
> Please help.
>
>
> --
> Tigran M. Abramyan, Ph.D.
> Postdoctoral Fellow, Computational Biophysics & Molecular Design
> Center for Integrative Chemical Biology and Drug Discovery
> Eshelman School of Pharmacy
> The University of North Carolina at Chapel Hill
> Chapel Hill, NC 27599-7363
>
> --
> 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.
>