General UM tips

CMS helpdesk

new cms helpdesk website

The NCAS UM tutorial

Check out a test suite

Suite u- should run on archer 2 with a swap of the charging code and user name

Rosie go search for suite and make a copy

cd <suite-name> rose edit

make required edits user name and charging code and run

Running a Suite

in a suite dir run:

rose suite-run --new rose suite-run --restart

viewing a running suites

rose sgc

Making ancils

Xancil

seg faults: ulimit -Ss 40000

ncview

Cannot read xconv converted ancils. e.g. qrparam.orog.nc will only be readable via xconv

Gotchas

Trouble shutting down a suite

ps -flu <username> | grep <suite-id>

then you can kill -9 <PID> the process that is still running preventing the suite from shutting down

Debugging tips

if the job.err messages are completely unhelpful there’s a number of things you can do.

  1. Try running: addr2line --exe=</path/too/executable> <address>

  2. Rerun in debug mode

in site/<HOST_HPC>/suite-adds.rc add

module load atp
export ATP_ENABLED=1

to the HOST_HPC init-script

AND

switching on extra output by

um > env > runtime controls > atmosphere only set PRINT_STATUS to “Extra diagnostic messages”

a similar thing can be done if your issue is in recon step

Upgrading a suites

cd [your suite]
rose macro --fix
cd app/um
rose app-upgrade vn12.0 (or whatever)
cd …/fcm_make

High Resolution Specific Notes

The Regional Area Science settings e.g. RA2 are scale aware.

For sub-km simulations the RMED team have a useful wiki page of settings (time steps etc) and adjustments that may be needed.

Troubleshooting

Ancillary generation issues

VEGFRAC failing

Generating subkm vegfrac ancillaries might fail in their interpolating. Its unclear why this is but you can use the fields from the parent nest where necessary. A neat way of doing this is to create exp1 config for your failing nest.

I’d just set the soil necessary ancillaries to source=1 (allow recon to decide) for the failing nest’s reconfiguration. This will then mean it picks up the parent nest ancillary data, which is more than good enough given the errors/approximations in the source data.

Assuming you are using the same science configuration for all resolutions you can copy to that configuration to the expt1. i.e. copy app/um/opt/rose-app-ra2t.conf (or whatever config you’re using) to app/um/opt/rose-app-expt1.conf .

Then add the ancil code from the app/um/rose-app.conf to the end of expt1.conf file, so for the latest version of the suite this is

[namelist:items(4c515841)]
ancilfilename='$UM_ANCIL_VEGFRAC_DIR/$UM_ANCIL_VEGFRAC_FILE'
domain=1
!!interval=1
l_ignore_ancil_grid_check=.true.
!!netcdf_varname='unset'
!!period=1
source=1
stash_req=216
update_anc=.false.
!!user_prog_ancil_stash_req=
!!user_prog_rconst=0.0

[namelist:items(884a098e)]
ancilfilename='$UM_ANCIL_SOIL_DIR/$UM_ANCIL_SOIL_FILE'
domain=1
!!interval=1
l_ignore_ancil_grid_check=.false.
!!netcdf_varname=10*'unset'
!!period=1
source=1
stash_req=40,41,43,44,46,47,48,207,220,223
update_anc=.false.
!!user_prog_ancil_stash_req=
!!user_prog_rconst=0.0

NB source=2 is edited to source=1

Then in the affected resolution configuration window make sure your change the ra2t or whatever in the drop down menu to expt1.

Domain Size

At high Resolution you may find your domain size is larger than the max sizes check for in Defensive checks

You will need to checkout a copy of the source code and edit

src/control/misc/atmos_max_sizes.F90

NB check the source you’re checking out matches you might need a branch of a branch!

Rebuild your executables and a larger domain will now be allowed