[MITgcm-support] MNC pickup files

Baylor Fox-Kemper baylor at MIT.EDU
Tue Aug 2 13:02:37 EDT 2005


Hi Jason,
   Depending on what version of the model you are using, you may need to 
delete the .0000. from the name of each pickup file.  Previous versions 
didn't like that naming.  I.e., change the pickups to:

> pickup.0000004000.000001.nc

Then it should (may?) work.
    -Baylor

On Aug 2, 2005, at 12:40 PM, Jason Goodman wrote:

> Is there any special trick to restarting MITGCM using MNC-created 
> NetCDF pickup files?  I copied the files from the MNC output 
> directories into the main run directory:
>
> -rwx------   1 goodmanj  admin  14562188 Aug  2 11:06 
> pickup.0000004000.0000.000001.nc*
> -rwx------   1 goodmanj  admin  14562188 Aug  2 11:06 
> pickup.0000004000.0000.000002.nc*
> -rwx------   1 goodmanj  admin  14562188 Aug  2 11:06 
> pickup.0000004000.0000.000003.nc*
> -rwx------   1 goodmanj  admin  14562188 Aug  2 11:06 
> pickup.0000004000.0000.000004.nc*
> -rwx------   1 goodmanj  admin  14562188 Aug  2 11:06 
> pickup.0000004000.0000.000005.nc*
> ...19 more...
>
> I and am using this data.mnc file:
>
> &MNC_01
> # mnc_echo_gvtypes=.FALSE.,
> mnc_use_indir=.FALSE.,
> mnc_use_outdir=.TRUE.,
> mnc_outdir_str='output_',
> mnc_outdir_date=.FALSE.,
> # monitor_mnc=.FALSE.,
> snapshot_mnc=.TRUE.,
> timeave_mnc=.FALSE.,
> pickup_write_mnc=.TRUE.,
> pickup_read_mnc=.TRUE.,
> &
>
> Timestepping in the "data" file is as follows:
>
> &PARM03
> nIter0=4000,
> nTimeSteps=5000,
> deltaT=2454,
> abEps=0.1,
> pChkptFreq=2454000,
> chkptFreq=1227000,
> dumpFreq=61350,
> monitorFreq=12270,
> &
>
> The model run fails with 24 errors like this:
> PGFIO-F-219/formatted write/internal file/attempt to read/write past 
> end of record.
> In source file MNC_CW_READWRITE_RL.f, at line number 6471
>
> 24-processor Linux cluster build, using Infiniband, MVAPICH, PGIF90 
> and GCC.
>
> I've noticed that the "ckptA" pickup file is twice the size of the 
> dated permanent checkpoints, and includes two timesteps worth of data.
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list