Failed to run GC5 with Ozone redistribution on ARCHER2

[xuzhang]

Hello,

I tried to run the GC5 with Ozone redistribution on ARCHER2. I have copied the trunk of u-dk922 to my account (as u-dv291) and then made the changes following the full instructions in below to enable the Ozone Redistribution with pp_transfer on:

https://cms.ncas.ac.uk/unified-model/configurations/ga9.0-gc5.0/

however, the run crashed on “fcm_make_pp_jasmin” because “make mirror” failed. I reran it in another day, but it still cannot go through this. could you help check this?

thank you

Xu

[AnnetteOsprey]

Hi Xu,

Have you set up you ssh connection from puma2 to Jasmin? See: Migration of UM Data to JASMIN Elastic Tape

If you have done this previously, can you check it is still working, e.g. can you connect on the puma2 command-line without a passphrase or password:

ssh sci-vm-03

Note that this step is only required if you want data automatically migrated to Jasmin ET via JDMA. If you don’t need this, you can switch it off in the suite and it will not run the fcm_make_pp_jasmin task.

Annette

[xuzhang]

Hi Annette

Thanks for your help.

After I switched it off, no such error message anymore.

Now it fails in the coupled job, because the job cannot find some essential files (see attached) to proceed. I guess I need to update the corresponding path in my SUITE to solve this error, right? if so, this is not listed in the webpage.

Best,

Xu

[AnnetteOsprey]

Hi Xu,

There is an error in the path in the site/archer2.rc file.

Change it to be:

    [[COUPLE_RESOURCE]]
        inherit = HPC
        script = "source $UMDIR/python/venvs/GC5_venv/bin/activate; export PYTHONPATH=$PYTHONPATH:$CYLC_WORKFLOW_SHARE_DIR/fcm_make_drivers/extract/drivers/Coupled_Drivers; {{TASK_RUN_COMMAND}} {{NON_RIGOROUS_PATH}}"

I will get Dan to update the standard workflows.

Annette

[xuzhang]

Hi Annette

It works well now. many thanks again for your help.

Cheers

Xu

[xuzhang]

Hi Annette

After finishing the first 2 months, the job failed in “retrieve_ozone” in the first month of the next year.

According to the error message, it appears that the source path to ozone ancillary files might be wrong.

Could you please check this?

Many thanks,

Xu

[AnnetteOsprey]

Hi Xu,

I have spotted the issue. There are some dodgy variable names which means the paths are not being set. In ozone-redistribution.rc update these lines:

    [[OZONE]]
        [[[environment]]]
{% set UM_INITIAL_OZONE_ANCIL = 'set_by_install_ancil' if OZONE_ANCIL_INITIAL is not defined %}
{% set UM_OROG_ANCIL = 'set_by_install_ancil' if OROG_ANCIL is not defined %}

The GC5 ozone redistribution scheme is quite complicated and I have done a lot of work to get it to work properly on Archer2. It should work, but we seem to have missed a couple of things when merging in the changes to the various workflows.

So thank you for reporting these issues. Do let me know whether it works now, and I will get these fixed in the base suites.

Thanks,
Annette

[xuzhang]

Hi Annette

thank you for porting the ozone suite on Archer2 and for helping me with this.

Till now the job has run for 1 year and 5 months smoothly after modifying those lines.

Best,
Xu