[MITgcm-devel] yesterday changes in global_ocean.cs32x15.viscA4

[Martin Losch]

Hi Jean-Michel,

here’s the content of the directory and the error message:
stan1:tr_run.viscA4> ls -l pickup*
lrwxrwxrwx 1 mlosch CLIDYN      26 May 13  2015 pickup.0000072000 -> ../input/pickup.0000072000
lrwxrwxrwx 1 mlosch CLIDYN      31 May 13  2015 pickup.0000072000.meta -> ../input/pickup.0000072000.meta
lrwxrwxrwx 1 mlosch CLIDYN      17 May 13  2015 pickup.0000086400 -> pickup.0000072000
lrwxrwxrwx 1 mlosch CLIDYN      38 May 13  2015 pickup.0000086400.meta -> ../input.viscA4/pickup.0000086400.meta
-rw-r--r-- 1 mlosch CLIDYN 7520256 May  8 06:46 pickup.ckptA.data
-rw-r--r-- 1 mlosch CLIDYN     379 May  8 06:46 pickup.ckptA.meta
-rw-r--r-- 1 mlosch CLIDYN 1474560 Jul 16  2006 pickup_nh.0000086400
lrwxrwxrwx 1 mlosch CLIDYN      39 May 14  2015 pickup_nh.0000086400.gz -> ../input.viscA4/pickup_nh.0000086400.gz
-rw-r--r-- 1 mlosch CLIDYN  737280 May  8 20:37 pickup_ph.0000086400
stan1:tr_run.viscA4> cat STDERR.0000 | tail -5
(PID.TID 0000.0001) WARNING >> READ_PICKUP:  no field-list found
(PID.TID 0000.0001) WARNING >>  try to read pickup as it used to be written
(PID.TID 0000.0001) WARNING >>  until checkpoint59i (2007 Oct 22)
(PID.TID 0000.0001) *** ERROR ***  MDS_READ_FIELD: filename: pickup_ph.0000086400.data
(PID.TID 0000.0001) *** ERROR ***  MDS_READ_FIELD: File does not exist

So pickup_ph.00000864000 is there, but I also have this error message:

run_clean skipped!
linkdata from dirs: input.viscA4 input
 ldir=input.viscA4: pickup_ph.0000086400.gz ;../input.viscA4/prepare_run: line 24: gunzip: command not found
 unzip files: pickup_ph.0000086400 ;
 ldir=input: eedata ; link files: from dir: ../../tutorial_held_suarez_cs/input
runmodel in global_ocean.cs32x15/tr_run.viscA4 ... failed (run: 1  end: 0 )
=> output from running in global_ocean.cs32x15/tr_run.viscA4 :
>  STOP ABNORMAL END: S/R MDS_READ_FIELD
> ./mitgcmuv(lang:f90): signal trap(SIGTERM: Software termination)
> stan-004: mpid: MPI process (universe = 0, rank = 1) terminated by exit(1)

I am not sure what’s happening, running the experiment alone with testreport ( like this:
./testreport -t global_ocean.cs32x15 -MPI 2 -of=../tools/build_options/SUPER-UX_SX-ACE_sxf90_awi  -small_f -runonly -command "mpirun -np TR_NPROC ./mitgcmuv” )
does not work (above error messages again), then I ran it directly (mpirun -np 2 ./mitgcmuv) which worked, then ran it again with testreport, and it worked again.
I am not sure how to debug this. Definitly “gunzip” does not work on the compute nodes of stan, maybe that’s the issue, because that’s the only place where gunzip is used?

Martin



> On 09 May 2016, at 16:14, Jean-Michel Campin <jmc at mit.edu> wrote:
> 
> Hi Martin,
> 
> It seems that the changes I made yesterday (most likely in 
> verification/global_ocean.cs32x15/input.viscA4/prepare_run
> ) are causing problems for the test you run on stan1.
> The file:  pickup_ph.0000086400 is not found but should have been
> gunzip when running "prepare_run"
> 
> When you have time, if you could take a look and tell me what is 
> wrong in "prepare_run", this would be nice.
> 
> Thanks,
> Jean-Michel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel