Currently trying to port an ensemble nesting suite at UM vn13.0 to ARCHER2 thanks to help from the RMED team at the MO. Since I’m not on Monsoon I have to rerun the MOGREPS-G members to drive the internal LAM, but I’m having trouble with the ENGL_UM_FCST site. It doesn’t seem to be accepting running in parallel and I’m not sure what settings to change. It gives the following error:
ereport_mod:???!!!???!!!??!!!???!!! ERROR ???!!!???!!!???!!!??!!! ereport_mod:? Error code: 100
ereport_mod:? Error from routine: UM_SHELL
ereport_mod:? Error message: UM started on 1 PEs but 720 asked for. Please adjust decomposition
ereport_mod:? Error from processor: 0
ereport_mod:? Error number: 0
It does run when I set dm_nproc=1,1. Maybe this is fine, but it seems a bit inefficient?
The suite I’m working on is u-do972.
Thanks in advance for any advice!
Best,
Fran
edit: sent message too early and also for some reason tagged monsoon; I’m very much not using monsoon haha
further context edit: I know that the engl app seems to be using the bin/um_forecast_wrapper file to call the um-atmos command instead of doing it directly, but even when I do swap it out and call um-atmos directly, it still won’t run in parallel
Not sure what exactly is going on with pytz but it seems to be installed under cray-python or postproc modules. Having a quick look it seems like the ensemble nesting suite uses it to check which model version is appropriate for the start date - important since analysis only exists for certain versions for certain dates (e.g. I’m using PS43 since I’m running a suite starting in October 2021, and PS43 is valid from 4/12/2019 to 4/5/2022 or something like that). I don’t know why but it doesn’t complain for me when I start the suite or run cylc validate.