Re: wrf_user_getvar segfault - for cape_2d

Do you have a way that you can watch the memory usage on your machine while you're running the script?
This may tell if you if you are running out of memory right when it fails.

--Mary

On Mar 19, 2012, at 7:52 PM, Will Thurston wrote:

> Hi Mary,
>
> I'm running ncl 6.0.0 and my uname is
>
> Linux 2.6.32.54-0.3-default #1 SMP 2012-01-27 17:38:56 +0100 x86_64
> x86_64 x86_64 GNU/Linux
>
> It's strange that it *does* work on copy of the file in which every
> alternate lat/lon point has been removed (eg ncks -O -d Time,0,1 -d
> west_east,,,2 -d south_north,,,2 -d south_north_stag,,,2 -d
> west_east_stag,,,2 wrfout_d03_2009-05-18_12:00:00.nc WRF_TST.nc). I'd
> be surprised if it's a memory/virtual memory problem though, as I'm on
> a cluster and giving it plenty of resources. Perhaps I should ask the
> cluster admins if it's a build problem they can fix?
>
> Will.
>
>
> On 14 March 2012 07:38, Mary Haley <haley@ucar.edu> wrote:
>> Hi Will,
>>
>> I was unable to reproduce the problem using NCL V6.0.0 and the WRF_DEBUG.nc file.
>>
>> What version of NCL do you have?
>>
>> --Mary
>>
>> On Mar 9, 2012, at 2:58 PM, Will Thurston wrote:
>>
>>> Hi again,
>>>
>>> I've uploaded a 2 timestep version of the file (on which getting cape
>>> segfaults for me) to the ftp site (put into incoming) created with
>>>
>>> ncks -O -d Time,0,1 wrfout_d03_2009-05-18_12:00:00 WRF_DEBUG.nc
>>>
>>> I'm running ncl 6.0.0 and my uname is
>>>
>>> Linux 2.6.32.54-0.3-default #1 SMP 2012-01-27 17:38:56 +0100 x86_64
>>> x86_64 x86_64 GNU/Linux
>>>
>>> Cheers,
>>> Will.
>>>
>>> On 3 March 2012 19:06, Will Thurston <willthurston@gmail.com> wrote:
>>>> Hi Dennis,
>>>>
>>>> I tried the script with a 2 timestep version of the wrfout file and it
>>>> still segfaulted when getting cape_2d. I also tried your ncks
>>>> "thinned-out" version (with just two timesteps and with all 24
>>>> timesteps) and both worked fine.
>>>>
>>>> It also works OK if I try it on a d01 or d02 file, so it seems like
>>>> the problem is to do with the number of points in the domain. I'd be
>>>> surprised if lack of memory was the problem though, I'm running on a
>>>> cluster and throwing plenty of mem/vmem at it.
>>>>
>>>> Cheers,
>>>> Will.
>>>>
>>>> On 3 March 2012 09:29, Dennis Shea <shea@ucar.edu> wrote:
>>>>> The original code was from wrfhelp [cc'd here].
>>>>>
>>>>> I ran you script with a standard WRF output file.
>>>>> I did add two printVarSummary statements.
>>>>> No problems. The output follows
>>>>>
>>>>>
>>>>> %> ncl cape_tst.ncl
>>>>>
>>>>> NCAR Command Language Version 6.0.0
>>>>>
>>>>> (0) Get dbz
>>>>>
>>>>> Variable: dbz
>>>>> Type: float
>>>>> Total Size: 1018464 bytes
>>>>> 254616 values
>>>>> Number of Dimensions: 3
>>>>> Dimensions and sizes: [Time | 24] x [south_north | 103] x [west_east |
>>>>> 103]
>>>>> Coordinates:
>>>>> Number Of Attributes: 6
>>>>> description : Max Reflectivity
>>>>> units : dBZ
>>>>> FieldType : 104
>>>>> MemoryOrder : XYZ
>>>>> stagger :
>>>>> coordinates : XLONG XLAT
>>>>> (0) Get CAPE
>>>>>
>>>>> Variable: cape
>>>>> Type: float
>>>>> Total Size: 4073856 bytes
>>>>> 1018464 values
>>>>> Number of Dimensions: 4
>>>>> Dimensions and sizes: [mcape_mcin_lcl_lfc | 4] x [Time | 24] x
>>>>> [south_north | 103] x [west_east | 103]
>>>>> Coordinates:
>>>>> Number Of Attributes: 5
>>>>> MemoryOrder : XY
>>>>> description : mcape ; mcin ; lcl ; lfc
>>>>> FieldType : 104
>>>>> stagger :
>>>>> coordinates : XLONG XLAT
>>>>>
>>>>> Try your script with (say) 2 time steps.
>>>>>
>>>>> Also, for debugging you may wish to do something like the following
>>>>>
>>>>> $> ncks -O -d Time,0,1 -d west_east,,,2 -d south_north,,,2 -d
>>>>> south_north_stag,,,2 -d west_east_stag,,,2 wrfout_d03_2009-05-18_12:00:00.nc
>>>>> WRF_TST.nc
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 03/02/2012 02:07 PM, Will Thurston wrote:
>>>>>>
>>>>>> Hi there,
>>>>>>
>>>>>> I've had problems with ncl seg faulting when using wrf_user_getvar to
>>>>>> calculate/get cape from some wrfout files, attached is a pared down
>>>>>> script.
>>>>>>
>>>>>> I've put the get dbz and print statements in to show (myself) that ncl
>>>>>> can calculate other diagnostic from 3D fields, and it is the cape part
>>>>>> it falls over on.
>>>>>>
>>>>>> The input file is about 22 Gb and has dimensions of 24 x 27 x 723 x
>>>>>> 732. This guess this is quite large so I also tried getting cape for
>>>>>> just 1 timestep, rather than the whole file at once, but it still seg
>>>>>> faults.
>>>>>>
>>>>>> ncl -V
>>>>>> 6.0.0
>>>>>>
>>>>>> Cheers,
>>>>>> Will.
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> ncl-talk mailing list
>>>>>> List instructions, subscriber options, unsubscribe:
>>>>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>>>>
>>>>>
>>> _______________________________________________
>>> ncl-talk mailing list
>>> List instructions, subscriber options, unsubscribe:
>>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Mar 20 09:39:57 2012