wrf.exe error while executing

[muzafar]

I am trying to execute wrf.exe and I am getting the below error

*** stack smashing detected ***: terminated

Program received signal SIGABRT: Process abort signal.

I am attaching the rsl.* files as well as namelist.input file.

@kwerner Sir here is a new link with all the required details

I want to run wrf for the dates mentioned in the namelist.input

 

[kwerner]

@muzafar,
Now that I look at this, I feel as though the issue could potentially still be related to the input data you mentioned in this post. The reason I say this is because the model stops immediately. Can you look through the variables in your wrfinput_d01 file to make sure everything looks okay? Another test you could do is to create the same simulation, but using another generic data type (such as GFS). If that runs, they you'll know it's still a data issue.

 

[muzafar]

@muzafar,
Now that I look at this, I feel as though the issue could potentially still be related to the input data you mentioned in this post. The reason I say this is because the model stops immediately. Can you look through the variables in your wrfinput_d01 file to make sure everything looks okay? Another test you could do is to create the same simulation, but using another generic data type (such as GFS). If that runs, they you'll know it's still a data issue.

I checked the file and it looks fine to me. I am attaching the txt files of the variables and the data values present in wrfinput_d01.

 

[muzafar]

I checked the file and it looks fine to me. I am attaching the txt files of the variables and the data values present in wrfinput_d01.

@kwerner Did I pass the correct information in the above files or there is something wrong in that

I downloaded the GFS data from

Index of /pub/data/nccf/com/gfs/prod

 

[kwerner]

I need to apologize - I got this post mixed up with another user's post, who was converting their own data. Please ignore my response above.

Can you send me your wrfinput_d01 and wrfbdy_d01 files? I would like to run a quick test using your exact files and namelist. The files will likely be too large to attach here, so if you don't have another method of getting them to me, see the home page of this forum for information on sharing large files. Thanks!

 

[muzafar]

I need to apologize - I got this post mixed up with another user's post, who was converting their own data. Please ignore my response above.

Can you send me your wrfinput_d01 and wrfbdy_d01 files? I would like to run a quick test using your exact files and namelist. The files will likely be too large to attach here, so if you don't have another method of getting them to me, see the home page of this forum for information on sharing large files. Thanks!

I have uploaded the two required files in a zip file named as wrffiles.zip to the Nextcloud portal.

Forum Uploads

Nextcloud - a safe home for all your data

nextcloud.mmm.ucar.edu

 

[muzafar]

@kwerner @Whatheway I have an analysis code written for wrf output file and I need a wrfout file with more than 50 timesteps of data. Is there anyway possible to get access to such an output file of WRF for now. I will keep trying in the meantime to fix the WRF execution with your help.

 

[Deleted member 3607]

@kwerner @Whatheway I have an analysis code written for wrf output file and I need a wrfout file with more than 50 timesteps of data. Is there anyway possible to get access to such an output file of WRF for now. I will keep trying in the meantime to fix the WRF execution with your help.

I am away from the computer for about a week due to illness. Will try to answer when back.

 

[kwerner]

@muzafar
I believe the issue may be that your time_step is too large. The value shouldn't be more than 6xDX, so you should reduce it. I used 135 and it ran to completion. Can you try that and see if it makes a difference?

Regarding a file with 50 timesteps of data, we don't have anything like that lying around. I apologize!

 

[muzafar]

@muzafar
I believe the issue may be that your time_step is too large. The value shouldn't be more than 6xDX, so you should reduce it. I used 135 and it ran to completion. Can you try that and see if it makes a difference?

Regarding a file with 50 timesteps of data, we don't have anything like that lying around. I apologize!

My DX 24000. I don't see how 6xDX is 135. I tried with 135 but still not able to run

 

[kwerner]

6xDX, where DX=24km is not 135. It's obviously 144, but in an effort to have the time-step (which is in seconds) match up with the history output time, which you've set to 180 (which is in minutes), I used a value that is less than 144. I chose 135 because it's evenly divisible by 15 seconds. It does not have to be done this way, but it's convenient.

As for your failure - because I was able to run without any problems when using your exact wrfbdy and wrfinput files, along with your namelist.input file, the problem likely has something to do with your specific system or environment. The issue could be that you don't have enough disk space to write the files, or that you don't have permission to write files to that directory. I would recommend contacting a systems administrator at your institution that can help you figure out what's going on. Let us know if you figure it out!

 

[muzafar]

6xDX, where DX=24km is not 135. It's obviously 144, but in an effort to have the time-step (which is in seconds) match up with the history output time, which you've set to 180 (which is in minutes), I used a value that is less than 144. I chose 135 because it's evenly divisible by 15 seconds. It does not have to be done this way, but it's convenient.

As for your failure - because I was able to run without any problems when using your exact wrfbdy and wrfinput files, along with your namelist.input file, the problem likely has something to do with your specific system or environment. The issue could be that you don't have enough disk space to write the files, or that you don't have permission to write files to that directory. I would recommend contacting a systems administrator at your institution that can help you figure out what's going on. Let us know if you figure it out!

I was executing with 8 mpi processes wrf and it still wasn't working
But then I ran it with only 1 process and it worked.
I was able to create the output file. But I am still not clear why it worked with 1 processes but not with more than 1 processes.
Is there any parameter for number of processes which I need to change.

 

[kwerner]

I'm glad you found a way to get it to work! This could potentially be an MPI issue, since when running with a single processor, it's not using multiple cores. If you're curious about that, you could test to see if 2 processors works. If not, then it's likely got something to do with your MPI installation and if you want to be able to use more processors, I'd recommend speaking to a systems administrator at your institution for help. There isn't anything that needs to be modified (parameter-wise) in the namelist or anything. It is possible to have too many or too few processors, depending on several factors, such as domain size, physics, etc., and you can read more about that here. But since I was able to run your case with 8 processors and it ran to completion, it seems to be an issue specific to your environment.

 

[muzafar]

I'm glad you found a way to get it to work! This could potentially be an MPI issue, since when running with a single processor, it's not using multiple cores. If you're curious about that, you could test to see if 2 processors works. If not, then it's likely got something to do with your MPI installation and if you want to be able to use more processors, I'd recommend speaking to a systems administrator at your institution for help. There isn't anything that needs to be modified (parameter-wise) in the namelist or anything. It is possible to have too many or too few processors, depending on several factors, such as domain size, physics, etc., and you can read more about that here. But since I was able to run your case with 8 processors and it ran to completion, it seems to be an issue specific to your environment.

Thank you. I will try to fix the MPI issue

 

[caralam]

Hi, muzafar,

I met the same problem as you mentioned above. I could provide a few more details.

I also ran wrf.exe with MPI, which worked well with my customized setting of &physics.

But when I turned to the default setting of &physics with "CONUS", I met the same error flagging as you reported.

I supposed the "CONUS" might not work well with our cases.

I hope this helps you.

 

[Gorrono]

Hi @muzafar ,
Thanks for posting your problem. After reading your thread, I think we are facing a similar issue.
I succesfully compiled with dmpar and nesting 1 and when I run wrf for a simple em-les test case I can only do it as ./wrf.exe or mpirun -np 1 ./wrf.exe

I tested and confirmed that mpirun works but as @kwerner there must be something else related to the way mpi processes are built with our system.
For info, WRF runs on a HP DL380P G9 server with 2 x Intel Xeon E5-2690 V4 (14 cores) 426 and 128GB RAM. I also attach the namelist and the error file given by one of the mpi processes.

I am investigating potential causes but no success yet. Please, if you find the issue after this time or anybody else has any idea, I would absolutely appreciate.

Thanks

 

[Gorrono]

Just wanted to let you know that in my case I found the issue
I have installed mpich 4.2.1 rather than mpich 3.0.4
In my case, I just followed the detailed installation instructions and download the version given on the link.
However, it is likely that the mpich version is old enough not to work on many systems (including mine)