New twist on the "program wrf: error opening wrfinput_d01 for reading ierr= -1021" real.exe error

[regclimo]

Hi all,

There have a few posts on this issue, but I'm in the unusual position that it seems to have been triggered by changing the land-cover specification. I've been using the same namelist.wps and namelist.input, but between yesterday and today I modified the GEOGRID.TBL file to match GEOGRID.TBL.ARW_LCZ (both versions attached; they have .txt suffixes only for upload purposes). Yesterday real.exe and wrf.exe completed as expected and I verified the final netcdf output files, but today I quickly get an error as shown in rsl.error.0000 and real.log. However, interestingly, the met_em output files produced today seem fine, so the error seems to be in real.exe specifically. To be absolutely sure nothing had changed on the cluster, I also went back and confirmed that everything still works when I punch in yesterday's settings.

What might I need to do to get real.exe and WRF to run successfully with these different geogrid parameters?

Thanks,
Colin

 

[kwerner]

Hi,
The error message in your log is:

-------------- FATAL CALLED ---------------
FATAL CALLED FROM FILE:  <stdin>  LINE:      70
 program wrf: error opening wrfinput_d01 for reading ierr=       -1021
-------------------------------------------

1) Are you able to view/open the wrfinput_d01 with something like ncview or ncdump?

2) For the case when you use the LCZ landuse, will you attach the rsl files from when you ran real.exe? You may have to run it again if you don't still have them. Please package all the rsl* files into a single *.tar file and just attach that.

3) Will you also issue the following in the WRF running directory, and then also attach the ls.txt file?

ls -ls >& ls.txt

Thanks!

 

[regclimo]

Thanks -- I should have explained better. No wrfinput files are created at all. The rsl.out.0000 file is attached; I looked through all the rsl.error and rsl.out files, and these are exactly the same (16 parallel processes hitting the same error), so not too useful for troubleshooting. The ls.txt is also attached.

 

[kwerner]

Thanks for the explanation. The new rsl file you sent is still for the wrf.exe run. Can you send me the rsl* files for the real.exe process? Even if they all seem the same, please package all of them together and send that packaged file. Thanks!

 

[regclimo]

It seems that the rsl* files for real.exe are all overwritten as soon as wrf.exe commences? I modified my batch script to stop after real.exe, examined the rsl* files, and found that I had a mismatch in the num_land_cat specification (num_land_cat in namelist.input needed to be 61). I resolved this, but now have the problem that WRF seems to be crashing early in the run without leaving behind many clues. Various files from the run are attached here -- the wrfinput* files are too large to include, but do look OK this time around.
Thanks again, Colin

 

[kwerner]

Hi,
I am so glad to hear that you resolved the original issue. To answer your question about the rsl files - yes, the files have the same names, so real.exe will run and print out however many rsl files you requested based on the number of processors you used to run real, and then when you run wrf, it will print out the number of files per # of processors you requested to use to run wrf.exe. So the real.exe rsl* files will be overwritten unless you move them elsewhere.

Since your new issue is different than the original problem you had here, will you please start a new thread with the new problem? Explain the error you're seeing and attach the appropriate files. This will keep each thread specific to one topic, which makes them easier to read in the future. Thanks!