list-directed I/O syntax error

[kgurer777]

Hello,

I downloaded MPAS-Model-8.0, compiled init_atmosphere, generated static, init and sfc_update files. However, after compiling atmosphere_model, I am running into list-directed I/O syntax error (please see out.mpas_run48705 file). I re-compiled the atmosphere_model in debug mode and ran it via 'debg atmosphere_model' (I don't know if running 'dbg atmosphere_model' is the proper way to debug the atmosphere_model. The user documentation and google search don't have that specific information on running MPAS in debug mode.). But I couldn't find the error. Howver, based on the error output message, the model complains about list-directed input/output at line 1552 in src/mpas_timekeeping.F file. While there are changes in MPAS-8.0 compared to previous version 7.3, I don't see how these lines could cause the error. Can you guide me to resolve this error?

Also, just as a side note, extrapolation of temperature that is set in namelist.init_atmosphere appears not yet implemented and atmosphere model looks for "config_len_displ" in namelist.atmosphere file.

Thank you.

Kemal.

 

[mgduda]

When compiling MPAS v8.0, are you using the PIO library or the SMIOL library? If the PIO library, which version have you used?

What messages in your log.init_atmosphere.0000.out file are you seeing that would indicate that the temperature extrapolation option hasn't been implemented?

Regarding the config_len_disp parameter, you'll need to use new mesh files (which are available on the mesh download page) if you'd like to omit this parameter from your namelist; otherwise, if you're using old mesh files, you'll need to supply a value for config_len_disp. For a longer explanation, please see the description of MPAS-Dev/MPAS-Model PR #1079.

 

[kgurer777]

Hello Michael,

I wanted to generate the errors with output files to demonstrate the issue for you, so it took a while to get back to you. I recompiled MPAS for init_atmosphere_model and atmosphere_model. I attached the log files as well as the namelist and batch files to submit both init_atmosphere_model and atmosphere_model.

I use a 3-48km variable mesh file that I had gotten from you about two years ago (x4x5.794626.tar.gz having 634mb size), which is not within the list of meshes available in mesh download page. I also use 3-60 km variable mesh available in the download page. For the sake of showing errors that I am coming up, I re-downloaded 3-60 km mesh and used in generating input files and running MPAS (see "x20.835586.roatated.grid.mesh_resolution.png" file).

The mesh files were originally centered over Atlantic Ocean, but I rotated it to center it over California as you can see from the plot file. My previous simulations with versions 6.x and 7.0 had run fine without any issues. The results shown here are from 3-60 km mesh only.

Here are the summaries of errors:
- Compilation with SMIO gives errors (see "save.log.init_atmosphere.0000.err" file) although it creates init executables (see save.out.make.init_atmosphere.with.SMIOL" file). Since I couldn't create initialization files, I couldn't check for atmosphere run.
- I used pio2 version 5_8 (see "save.out.make.init_atmosphere" file).
- After getting init compilation successful with pio2, I can get static file fine but it gives error on "extrap_type = 2 not implemented" error (see "save.log.init_atmosphere.err.2" file). Once I change the extrapolation from "lapse_rate" linear, the model produces init and sfc_update files without any additional issues.
- I clean init_atmosphere and create atmosphere executable fine (see "out.save.out.make.atmosphere" file). However, when I try to run it, it first complains about the "config_len_disp" (see save.log.atmosphere.0000.err" file). After putting it into namelist.atmosphere (see "namelist.atmosphere" file), this time it stops with list_directed I/O error (see "save.out.mpas_run48785" file) albeit without letting me know where it is running into issue (see also "save.log.atmosphere.0000.out" file).

Thank you for looking into this error.

Kemal.

 

[kgurer777]

I cannot see all attachments. If they are not there, please let me know. - Kemal.

 

[mgduda]

I'm only seeing the attachment with the mesh plot. Could you give another try at attaching the log files?

 

[kgurer777]

Here it is. Hopefully this will work. Thanks.

 

[kgurer777]

Hello Michael,

I just wanted to report the results of latest run with SMIOL and with PIO2 libraries and without any mid-processes such as rotating the mesh and creating graph.info partition with the ones that I typically use based on the number of nodes that I have, Instead, I just used 36 CPU partition that came with the system.

I re-downloaded MPAS-Model-8.0.0.tar.gz file, untarred, compiled and ran again with untarred x20.835586.tar.gz file and without rotating it to center over California this morning. Previous error that I was mentioning about SMIOL library, extrapolation option and config length are not seen this time. I don't know why that was the case earlier. This time, the results of both SMIOL and PIO2 options are the same such that I don't get the SMIOL error message and it doesn't complain about config length and extrapolation option. However, the list-directed-I/O error is seen in both runs. Please see the zip files of the results of both options as attached to this message. I cannot figure out where MPAS is trying to read names/values and complaining about not reading properly.

Thank you for your help Michael.

Kemal.

 

[kgurer777]

Hi Michael,

Please disregard the previous emails that I sent. I solved the problem partially. When I create input files using "io_type="pnetcdf,cdf5", then the model doesn't complain about I/O error and it runs. However, this time, I have to set "config_sst_update = false" and "config_sstdiurn_update = false" in namelist.atmosphere file. When I don't use "io_type="pnetcdf,cdf5" option in streams.init_atmosphere, then I can update SST in namelist.atmosphere. I didn't have this problem with MPAS.7.0 but I have this problem with versions 7.3 and 8.0. Could you let me know how I can resolve this problem of updating SST without I/O error? Thank you.

Kemal.

 

[kgurer777]

Hello again Michael.

I figured out that "input_interval" for SST in streams.atmosphere file also has to be set to the output interval of sfc/sst file. Then the model runs with SST update. Sorry about occupying you with the thought of what could be wrong. With regards,

Kemal.

 

[mgduda]

Apologies for the long delay in following up here! I just wanted to check at this point that you're able to run with all of the options that you'd like to use (rotated mesh refinement, SST update, etc.)? In any case, no worries about asking questions -- it's through these sorts of threads that we're often able to identify subtle bugs in the code that didn't turn up in our own testing.

 

[kgurer777]

Hi Michael.

Yes, I was able to resolve the issues in running MPAS version 8.0. Thanks for checking back with me.

Kemal.