Re: [swanmodel-users] mpi run of SWAN 41.31 is not able to merge netcdf output files
Brought to you by:
mzijlema
Menu
▾
▴
Re: [swanmodel-users] mpi run of SWAN 41.31 is not able to merge netcdf output files
|
From: Javi R. <ja...@pr...> - 2021-07-05 20:23:11
|
Hi you all,
Trying to find out what is wrong with those circumstances can be very tricky. You are running a newer version of Swan also with a newer version of the NetCDF library, so we don't know which change makes the issue arise. Remember, the official Swan documentation asks about using NetCDF 4.1.6 or similar, and the library is facing a pretty dynamic stage of development.
Also, Swan is not able to export all the variables to the NetCDF format, only a certain ones really work. I remember I had certain issues with recent versions of the NetCDF library (4.6.3 I guess), so I finally ended up exporting my model results to a Matlab (.mat) file and converted it to .nc using a Python script. Voilà, zero issues.
My recommendation for now is that you must try to change your NetCdF libraries back to an older version, and check if Swan 41.31 works or not (There's no need to change the Intel Fortran compiler). If it doesn't, I guess you must use your previous environment and wait for either the Swan team or the NetCDF team to fix those issues in future releases.
Regards,
Javi
-------- Missatge Original --------
A 1 jul. 2021, 10.26, ANGEL MARTINEZ FERRER va escriure:
> Hi Javier,
>
> Yes, SWAN generates these files:
>
> -rw-r--r--. 1 pn16 pn 24104 Jul 1 07:35 hs_2021062300.nc
> -rw-r--r--. 1 pn16 pn 139128 Jul 1 07:35 hs_2021062300.nc-001
> -rw-r--r--. 1 pn16 pn 137088 Jul 1 07:35 hs_2021062300.nc-002
> -rw-r--r--. 1 pn16 pn 102816 Jul 1 07:35 hs_2021062300.nc-003
> -rw-r--r--. 1 pn16 pn 111384 Jul 1 07:35 hs_2021062300.nc-004
> -rw-r--r--. 1 pn16 pn 100776 Jul 1 07:35 hs_2021062300.nc-005
> -rw-r--r--. 1 pn16 pn 100776 Jul 1 07:35 hs_2021062300.nc-006
> -rw-r--r--. 1 pn16 pn 95472 Jul 1 07:35 hs_2021062300.nc-007
> -rw-r--r--. 1 pn16 pn 183600 Jul 1 07:35 hs_2021062300.nc-008
> -rw-r--r--. 1 pn16 pn 95472 Jul 1 07:35 hs_2021062300.nc-009
> -rw-r--r--. 1 pn16 pn 97920 Jul 1 07:35 hs_2021062300.nc-010
> -rw-r--r--. 1 pn16 pn 97920 Jul 1 07:35 hs_2021062300.nc-011
> -rw-r--r--. 1 pn16 pn 100980 Jul 1 07:35 hs_2021062300.nc-012
>
> hs_2021062300.nc is the Hsig field and is the only proper netCDF file. It is attached to this mail.
>
> This is the BLOCK command in the INPUT file:
>
> FRAME 'mapa' xpfr=-0.50 ypfr=38.75 alpfr=0.0 xlenfr=.875 ylenfr=1.125 mxfr=84 myfr=108
>
> QUANTITY HS excv=-99.0
> BLOCK 'mapa' NOHEADER 'hs_2021062300.nc' LAY-OUT 1 HS OUTPUT tbegblk=20210623.000000 deltblk=3 HR
> OUTPUT BLOCK ndec=2 len=1400
> hs_2021062300.nc was supposed to contain 17 time fields, but only has one. The values for Hsig are only NaNs.
>
> This is its header:
>
> ncdump hs_2021062300.nc|more
> netcdf hs_2021062300 {
> dimensions:
> time = UNLIMITED ; // (1 currently)
> longitude = 85 ;
> latitude = 109 ;
> variables:
> int time(time) ;
> time:units = "seconds since 1970-01-01" ;
> time:calendar = "gregorian" ;
> time:standard_name = "time" ;
> time:long_name = "time" ;
> float longitude(longitude) ;
> longitude:units = "degrees_east" ;
> longitude:long_name = "longitude" ;
> longitude:standard_name = "longitude" ;
> float latitude(latitude) ;
> latitude:units = "degrees_north" ;
> latitude:long_name = "latitude" ;
> latitude:standard_name = "latitude" ;
> short hs(time, latitude, longitude) ;
> hs:units = "m" ;
> hs:standard_name = "sea_surface_wave_significant_height" ;
> hs:long_name = "hs" ;
> hs:scale_factor = 0.0007629627f ;
> hs:add_offset = 25.f ;
> hs:_FillValue = -32768s ;
>
> // global attributes:
> :Conventions = "CF-1.5" ;
> :History = "Created with agioncmd version 1.5" ;
> :Directional_convention = "cartesian" ;
> :project = "VALE2021062300" ;
> :run = "1" ;
> data:
>
> time = 1624406400 ;
>
> longitude = -0.5, -0.4895833,....
>
> You are right: the file may not be even opened. But why? Any help will be appreciated.
>
> Thanks,
>
> Ángel
>
> sha...@ca... escribió:
>
>> Hi Angle,
>>
>> I suddenly am experiencing a similar problem with SWASH that never happened before now.
>>
>> Were you able to find a solution?
>>
>> Shai
>>
>> From: Javi Rodriguez via swanmodel-users <swa...@li...>
>> Sent: Wednesday, June 30, 2021 6:20 PM
>> To: ANGEL MARTINEZ FERRER <ama...@ae...>
>> Cc: swa...@li...
>> Subject: Re: [swanmodel-users] mpi run of SWAN 41.31 is not able to merge netcdf output files
>>
>> Hi Angel,
>>
>> When you ran the model, did Swan generate any .nc files? It seems like the file to be written is not even opened...
>>
>> Regards,
>>
>> Javi
>>
>> ‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
>> On Wednesday, June 30th, 2021 at 14:25, ANGEL MARTINEZ FERRER <ama...@ae...> wrote:
>>
>>> Dear SWAN users,
>>>
>>> I encountered an error at the end of a mpi run of SWAN 41.31A
>>>
>>> This is the first execution in a new linux cluster with 1 node and 12 cpus. Everything runs fine, but at the end SWAN doesn't merge the netCDF files into one. It gives next message:
>>>
>>> NetCDF: Not a valid ID
>>>
>>> Compilation took into account netCDF location. When checking ldd with the swan executable, all libraries are known.
>>>
>>> The swaninit file has 99999 as the highest file reference number.
>>>
>>> This are the modules I loaded when compiling and executing SWAN:
>>>
>>> intel/2019
>>>
>>> impi/2019
>>>
>>> zlib/1.2.8
>>>
>>> hdf5/1.10.6
>>>
>>> netcdf/4.7.4
>>>
>>> SWAN ran OK in a former linux cluster with this modules:
>>>
>>> module load intel/2016
>>>
>>> module load impi/5.1
>>>
>>> module load netcdf/4.3.2
>>>
>>> Can you give any hint on this issue?
>>>
>>> Thanks a lot
>>
>> External e-mail, be judicious when opening attachments or links |