Running the jules suite with a different source code (cont...)

JULES
1

Hi,

A follow up to the task taken up to run the suite with a different source code. (Earlier topic was: Running an suite with a new jules source code)

  • There were a few namelist variables that are missing in the vn7.0_jules I am trying to use, so commented them and the RECON runs without an error. However, this is creating a very large job.out file (last one I deleted was 57 GB) and stopping all suites running.

  • I am showing below the namelist parameters that I changed (in rose-suite.conf):
    → &jules_vegetation:: l_sugar
    → & jules_vegetation::rd_model
    → &jules_soil_biogeochem::kaps_4pool=3.22e-7,9.65e-9,2.12e-8,6.43e-10,
    → &jules_latlon:: !!l_coord_latlon=.true. → not available in vn7.0
    → &pft_params:: → not present in vn7.0

    • !!sug_g0_io=
    • !!sug_grec_io=
    • sug_yg_io=

  • I have submitted a new simulation (u-knr47/run1) now. And I beileve the same is happening, where a very large job.out file is being genearted. This did not happen when running the same code with other source code (which is vn7.6).

Is there something that I am missing here?

The suite I am running now is u-knr47 (in /home/users/naren047/roses/) using the src code from /home/users/naren047/jules_src/vn7.0copy1_c4. The earlier source code which I used for running the suite was/home/users/naren047/jules_src/vn7.6_vcmax-jmax-diurnal.

Do let me know if any other information is required to inspect/resolve this issue.

Thanks and regards,
Narender

2

Hi Narender

Most of the space in job.out is taken up by printing arrays headed npp_l and anetl. These seem to come from print statements in the source files src/science/surface/leaf_jls_mod.F90 and src/science/surface/sf_stom_jls_mod.F90. I suspect that somebody added these for diagnostic purposes at some point and then forgot to delete them or comment them out. If you look in the two source files, you will see that there are rather a lot of print statements. Most of them are commented out, but not all. I suggest that you comment them all out unless you really need them.

Good luck
David

3

Hi David,

Thanks for the help. I have commented the print statements. Now the run does not create any large files that disrupts the simulations. However, it still does not succeed. I have checked the log files but could not figure out exactly what is causing this error and failing of RECON part of the suite.

Please help in pointing to the cause of the simulation failure.

Thanks

Narender

4

Hi Narender

It appears from job.err that the program is failing due to a segmentation fault in the subroutine snowtherm. This is usually a sign of an array index that’s outside the bounds of an array. Unless you have an idea of how this might be happening, the next step is to gather information on what exactly is going wrong. Unless you know how to run JULES in a debugger, I’d add some print statements to the source file src/science/snow/snowtherm_mod.F90. The first thing I’d try is to print the value of l after the statement l = surft_index(k). It’s indirect array indices like these that have the most scope for going wrong.

Good luck
David