[gmx-users] Re: g_energy: Energy names magic number mismatch

Mikhail Stukan mstukan at slb.com
Mon Dec 13 10:00:18 CET 2010


Justin,

Thanks a lot.
Concatenation seems to be the only solution, but it works fine. It look like there are some problems with the appending routine on BlueGene.

Mikhail

> Dear Justin and Berk,
>
> Thank you very much for the hint. Unfortunately, upgrade to 4.5.3 version doesn't solve the problem, diagnostic is still the same:
> ==============================
> Program g_energy_d, VERSION 4.5.3
>
> Source code file: ../../../src/gmxlib/enxio.c, line: 422
>
>
> Fatal error:
>
> Energy names magic number mismatch, this is not a GROMACS edr file For
> more information and tips for troubleshooting, please check the GROMACS
> website at http://www.gromacs.org/Documentation/Errors
> ===============================
>
> Do you think there could be any way to overcome this problem apart from running the job in parts?
>

Just to clarify, now that you've upgraded to 4.5.3, are you also running the
4.5.3 version of mdrun, and not just analyzing a 4.5.1 .edr file with 4.5.3?  I
believe there have been changes to the .edr file format over the course of
development.

There are several things to try to better diagnose what's going on.  If file
appending isn't working properly, that would be a major problem.  To diagnose:

1. What does gmxcheck tell you about the problematic .edr file?
2. Try running the jobs as separate intervals.  Check the integrity of each with
gmxcheck and concatenate them with eneconv.  Then use gmxcheck on this manually
concatenated .edr file.  If g_energy works on this file, then the problem is in
the appending routines, not the file format.

-Justin

> Many thanks in advance,
> Mikhail
>
>
>
>> Dear gromacs users,
>>
>>
>>
>> I am facing the following problem while running gromacs 4.5.1 at
>> BlueGene/P supercomputer.
>>
>>
>>
>> The job is submitted by the following script file:
>>
>>
>>
>> #!/bin/sh
>>
>> #
>>
>> # @ account_no          = xxx
>>
>> # @ job_name            = NVT1
>>
>> # @ job_type            = bluegene
>>
>> # @ output              = $(job_name).$(jobid).out
>>
>> # @ error               = $(job_name).$(jobid).err
>>
>> # @ environment         = COPY_ALL;
>>
>> # @ wall_clock_limit    = 24:00:00
>>
>> # @ notification        = always
>>
>> # @ bg_size             = 64
>>
>> # @ queue ?/mpirun -cwd $PWD  -mode VN -np 256 -exe ?/mdrun_bgp_d -args
>> "-deffnm nvt1 -dds 0.5 -cpi nvt1.cpt -append"
>>
>>
>>
>>
>>
>> And it is running perfectly fine. I can analyze the data using g_energy
>> etc. But if after the run is finished I increase the number of time
>> steps and submit run again or resubmit it after a crash (killed manually
>> or due to 24 hours wall clock limit), using the same script provided
>> above, the run itself looks fine, at least .log file looks as it should
>> but when I try to analyze .edr file with g_energy tool I obtain the
>> following error message:
>>
>> -------------------------------------------------------
>>
>> Program g_energy_d, VERSION 4.5.1
>>
>> Source code file: ../../../src/gmxlib/enxio.c, line: 409
>>
>>
>>
>> Fatal error:
>>
>> Energy names magic number mismatch, this is not a GROMACS edr file For
>> more information and tips for troubleshooting, please check the GROMACS
>> website at http://www.gromacs.org/Documentation/Errors
>>
>> -------------------------------------------------------
>>
>>
>>
>> I would really appreciate any hint on this subject.
>>
>
> This was a Gromacs bug in versions 4.5.1 and 4.5.2.  Please upgrade to version
> 4.5.3.
>
> http://www.gromacs.org/About_Gromacs/Release_Notes/Versions_4.5.x
>
> -Justin
>
>>
>> Many thanks in advance.
>>
>> Mikhail.







More information about the gromacs.org_gmx-users mailing list