Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

MPAS v8 with SMIOL presents error

M

mantovanijr

New member
Hi everybody! I successfully installed MPAS v8 without USE_PIO argument, then the MPAS was compiled with SMIOL I/O library.

However, when I run the init_atmosphere step to interpolating the atmospheric data I've got the error as follows:

** Attempting to bootstrap MPAS framework using stream: input
ERROR: SMIOLf_open_file failed with error -5

Only to your knowledge, I builded the MPAS v8 with the same libraries which I'd build MPAS v7.3 (No issues with MPAS v7.3 during any step of model run).

I've attached log files. The mesh x1.10242 (240 km uniform) was downloaded from MPAS meshes repository.

Anyone could help me to solve it? I would like to move from MPAS v7.3 to v8. Let me know if I should try MPAS v8 with PIO rather than SMIOL (I want use SMIOL since apparently it performs better as discussed during Joint WRF&MPAS workshop).
 

Attachments

  • log.init_atmosphere.0000.err.txt
    335 bytes · Views: 8
  • log.init_atmosphere.0000.out.txt
    2.2 KB · Views: 8
Your log.init_atmosphere.0000.out indicates that many variables are missing in your file x1.10242.static.nc.

Please issue the command:

ncdump -h x1.10242.static.nc > log

And send me your log file to take a look. Thanks.
 
Ming Chen said:
Your log.init_atmosphere.0000.out indicates that many variables are missing in your file x1.10242.static.nc.

Please issue the command:

ncdump -h x1.10242.static.nc > log

And send me your log file to take a look. Thanks.
Click to expand...
I appreciate the quickly reply.

When I type the command (ncdump -h x1.10242.static.nc > log) the terminal returns: ncdump: x1.10242.static.nc: NetCDF: Unknown file format.

I've downloaded the mesh from this URL (MPAS-Atmosphere mesh downloads) and uses tar -xzvf to untar the file.

Edited: I redo the procedure on my laptop and sent the result as attachment.
 

Attachments

  • log.txt
    12.2 KB · Views: 1
Last edited:
I just downloaded the same file from the link in your post. The file looks fine. See the log file I attached here.
Please try to re-download the file.
 

Attachments

  • log.txt
    12.2 KB · Views: 0
Ming Chen said:
I just downloaded the same file from the link in your post. The file looks fine. See the log file I attached here.
Please try to re-download the file.
Click to expand...
Note that I've edited my answer. I made it on my laptop and works fine. But, I am doing the procedure on the supercomputer.
 
Ming Chen I've moved the x1.10242.static.nc file that I've downloaded on my laptop to supercomputer using scp. I sent the job to run init_atmosphere and It works fine! I don't know why and what is causing this issue.. It's not good..
 
@mantovanijr Could you help me to understand the situation more clearly? You've originally tried running on your HPC cluster with the x1.10242 (static or grid?) file that you've downloaded from the mesh download page; but that resulted in the error you posted. However, after you've downloaded the static file on your laptop, then copied it to your HPC cluster, the init_atmosphere_model program is now working?
 
mgduda said:
@mantovanijr Could you help me to understand the situation more clearly? You've originally tried running on your HPC cluster with the x1.10242 (static or grid?) file that you've downloaded from the mesh download page; but that resulted in the error you posted. However, after you've downloaded the static file on your laptop, then copied it to your HPC cluster, the init_atmosphere_model program is now working?
Click to expand...
For sure, exactly this. When I download the x1.10242.static file direct from the HPC cluster, untar the file, and run init_atmosphere via submission job I got this error which you saw through this thread. I try to make the command ncdump -h x1.10242.static.nc > log) and the terminal returns: ncdump: x1.10242.static.nc: NetCDF: Unknown file format. So due to it, I decided to download the x1.10242.static file directly from my laptop, and the command worked. Then I think: Why not send the file to HPC cluster via scp? And then, all the rest works... including the atmosphere_model by now. Would be the tar -xzvf command or maybe wget the problem?
 
Thanks so much for following up, and that is puzzling indeed. I use 'wget' to download mesh files regularly without issue, and similarly for 'tar -xzvf ...'.

I wonder if it's possible that the download was somehow corrupted the first try on your HPC cluster? Just out of curiosity, do you get the same SHA256 hash for the x1.10242.static.nc file:
Code: 
$ openssl sha256 x1.10242.static.nc
SHA256(x1.10242.static.nc)= b380f5fb8b9835e5b1c3252f0e89c1d0b5b922daace1c607d89b85e40f3661d8

Also, it would be interesting to know if you re-download the static file on your HPC cluster with 'wget', would you be able to reproduce the original error from the init_atmosphere_model program?

There's of course no obligation to carry out any of these additional tests if everything is running, now; but, I'm rather interested to try to get to the bottom of the original issue in case there's anything we need to fix on our side.
 
mgduda said:
Thanks so much for following up, and that is puzzling indeed. I use 'wget' to download mesh files regularly without issue, and similarly for 'tar -xzvf ...'.

I wonder if it's possible that the download was somehow corrupted the first try on your HPC cluster? Just out of curiosity, do you get the same SHA256 hash for the x1.10242.static.nc file:
Code: 
$ openssl sha256 x1.10242.static.nc
SHA256(x1.10242.static.nc)= b380f5fb8b9835e5b1c3252f0e89c1d0b5b922daace1c607d89b85e40f3661d8

Also, it would be interesting to know if you re-download the static file on your HPC cluster with 'wget', would you be able to reproduce the original error from the init_atmosphere_model program?

There's of course no obligation to carry out any of these additional tests if everything is running, now; but, I'm rather interested to try to get to the bottom of the original issue in case there's anything we need to fix on our side.
Click to expand...
Sure thing, it's really interesting..
1. Well, I got the same SHA256 hash for the x1.10242.static.nc file that I've moved to HPC cluster (which was downloaded on my laptop).
2. After 1., I've excluded the file and I've downloaded the file directly from the HPC cluster via wget, made the command tar -xzvf.., and I've got the same as in 1. (SHA256(x1.10242.static.nc)= b380f5fb8b9835e5b1c3252f0e89c1d0b5b922daace1c607d89b85e40f3661d8)
3. After 2., I made some modifications in namelists and also a clean up in the directory (removed output files), then submitted the job and the job failed 6s after starting. I've got the same error than before (attached files in this thread). It's so puzzlingly.
4. Note that, the x1.10242.static.nc file was downloaded (step 3. above) in a directory below the directory where namelists and other files are placed (an 240km_uniform directory as in MPAS tutorial), and this file was symbolically linked via ln -sf. So, finally I've tried to move the file for the work directory and now both init_atmosphere as well as atmosphere_model works fine.
5. Maybe, Is strictly required that x1.10242.static.nc file be placed in the "work directory" rather than anyone else?
Thank you for the response. I would like to help as possible. I am from INPE in Brazil, and here we are working with MPAS.
 
You must log in or register to reply here.
Top