Error when testing nudging tools in CESM2.1.3

[Jerry]

Hi, everyone!
I try to test the nudging tools in directory 'components/cam/tools/nudging/Gen_Data/'.
And I chose ERAI as the target reanalysis product.
I edited the csh script according to the instructions in the README file and submitted it for execution.
Unfortunately, there was an error.
I have attached the edited csh script and the Slurm log file and hope this can provide some help.
Thanks!

 

[Jerry]

By the way, the NCL version I used is 6.6.2.

 

[peverley]

Hi Jerry,

I would recommend adjusting some of these paths. I don't believe the script will create directories, so if ./Config doesn't exist, for example, you will get errors.

# Set INPUT/OUTPUT/TMP directories,
#-----------------------------------
set NAMELIST = './Config/Config_makeIC-'$RUNNUM'.nl'
set MYLOGDIR = './LOG/LOG_001.'$RUNNUM'/'
set MYTMPDIR = './TMP/TMP_001.'$RUNNUM'/'
set MYOUTDIR = '/share/home/caost/scripts/output/'
set INPUTDIR = '/share/home/caost/scripts/erai/'

 

[Jerry]

Hi Jerry,

I would recommend adjusting some of these paths. I don't believe the script will create directories, so if ./Config doesn't exist, for example, you will get errors.

Hello Peverley,
Thanks for your reply. I have identified the statement causing the error in the csh script.
The error is related to the WRAPIT command, which is used to call functions from Fortran files.
But I still have no idea how to resolve the error. Maybe It is related to the NCL version.

 

[peverley]

Hi Jerry,

Did you by chance try updating those paths? When I ran it, I was getting an error that said it was about WRAPIT, but was actually a symptom of the path issue.

 

[Jerry]

Hi Jerry,

Did you by chance try updating those paths? When I ran it, I was getting an error that said it was about WRAPIT, but was actually a symptom of the path issue.

Wow, that is amazing. I will have a try!
Thank you for your kind attention.

 

[Jerry]

Hi Jerry,

Did you by chance try updating those paths? When I ran it, I was getting an error that said it was about WRAPIT, but was actually a symptom of the path issue.

Hi Peverley,
I have created several folders as required by the script; however, the task still failed a few seconds after submission.
The error still exists: *** buffer overflow detected ***: /share/apps/anaconda3/envs/NCtools/bin/wrapit77 terminate.
By the way, Can you tell me the Linux version, NCL version, and Fortran compiler you are using？

 

[peverley]

Can you include the full error output? Did it seem to get further?

My versions of Linux, NCL, Fortran compiler (intel)
SUSE Linux 5.14.21-150400.24.18-default
ncl/6.6.2
intel/2023.0.0

 

[Jerry]

Can you include the full error output? Did it seem to get further?

My versions of Linux, NCL, Fortran compiler (intel)
SUSE Linux 5.14.21-150400.24.18-default
ncl/6.6.2
intel/2023.0.0

The "slurm-6548.out.txt" attachment is the latest log file after I submit the script.
It was aborted after being submitted for a few seconds. This seems to be progress compared to being terminated immediately before.
The log file contains error messages about WRAPIT and the absence of some folders, but I guess the latter seems to depend on the former?

 

[Jerry]

Dear all,
The above issue has been resolved, and I appreciate everyone who provided suggestions. The error related to WRAPIT BUFFER OVERFLOW was ultimately traced to a subroutine named read_namelist in the MAKEIC.f90 file. The excessive number of parameters in this subroutine was identified as the cause of the error. By reducing the parameters to 31 or fewer (the reason for the critical threshold being 31 is not yet clear), the error was resolved. My own solution involved rewriting the read_namelist subroutine into two similar subroutines, each containing a subset of the original parameters, ensuring that the parameter count remained below the critical threshold.
I hope this can be helpful to anyone encountering a similar issue.

 

[zyc]

尊敬的各位：
上述问题已解决，感谢所有提供建议的人。与 WRAPIT BUFFER OVERFLOW 相关的错误最终被追踪到 MAKEIC.f90 文件中名为 read_namelist 的子例程。此子例程中的参数过多被确定为错误的原因。通过将参数减少到 31 或更少（临界阈值为 31 的原因尚不清楚），该错误已得到解决。我自己的解决方案包括将 read_namelist 子例程重写为两个相似的子例程，每个子例程都包含原始参数的子集，确保参数计数保持在临界阈值以下。
我希望这对遇到类似问题的任何人都有所帮助。

嗨，杰瑞！
我也在运行插值脚本，但我有一些错误似乎没有解决，如果您能帮助我，我将不胜感激。

 

[Jerry]

嗨，杰瑞！
我也在运行插值脚本，但我有一些错误似乎没有解决，如果您能帮助我，我将不胜感激。

hi, Yichao,
Based on the log files you provided, I think the error you encountered differs from mine. It seems that the required namelist file for executing the interpolation was not successfully created.
The target grid type I need is finite volume(fv), so I don't need to execute the file "ESMF_regridding.ncl".

 

[zyc]

Hi Jerry,
Oh, I see. Anyway, thank you for your help!