Opened 7 weeks ago

Last modified 9 days ago

#3056 reopened help

GRIB start dump

Reported by: mtodt Owned by: um_support
Component: UM Reconfiguration Keywords: GRIB ARCHER
Cc: Platform: ARCHER
UM Version: 10.6

Description

Hi

I am trying to re-run a suite on ARCHER that was originally run on MONSOON and used the start dump /data/d05/hadom/startdumps/era20c_1950010100.grib. My suite fails with following error message:
? Error code: 100
? Error from routine: UKMO_DECODE
? Error message: Attempting to read GRIB data, but GRIB_API ifdef not present
? in build. Please note the GRIB API is currently the only method
? for decoding GRIB

I've read previous tickets (2739 and 2835) and understand that at those points GRIB_API was not supported for UM versions <11. Is that still the case? If so, is there a non-GRIB version of this start dump available on ARCHER? (I will have to stick to my UM10.6 version for comparability.)

Many thanks in advance!

Cheers
Markus

Change History (20)

comment:1 Changed 7 weeks ago by willie

Hi Markus,

There are no UM start dumps in MASS prior to 2006. If you have access to the Monsoon run you could use the reconfigured start dump from that. What is the suite id?

Willie

comment:2 Changed 7 weeks ago by mtodt

Hi Willie

Thanks for the quick response! My suite is u-bn412, which is based on u-ak185. I don't have access to MONSOON, but could ask someone to get it for me. Where would I find the reconfigured dump file?

Cheers
Markus

comment:3 Changed 7 weeks ago by willie

Hi Markus,

I thought I would just run u-ak185 on Monsoon for you, but it is quite old and is not working now that the old exvmsrose computer has gone. The job was last run in Feb 2017 by Malcolm Roberts and some of the output, but not the reconfigured start dump, is archived in MASS. The earliest start dump is from moose:/crum/u-ak185/ada.file/ak185a.da19510101_00 (one year later than you asked for) and I have put this on ARCHER under /work/n02/n02/wmcginty. Please copy it to your own workspace if it is useful to you.

As for u-bn412, this may need some modification to get it to work on ARCHER. This is a 10.6 job, so my 11.1 branch used in previous tickets won't work.

I will continue to looks at this.

Willie

comment:4 Changed 7 weeks ago by mtodt

Hi Willie

Thanks a lot! I copied the dump file to my directory.

Cheers
Markus

comment:5 Changed 6 weeks ago by willie

Hi Markus,

I have gone back to the original Monsoon suite u-ak185 and upgraded it to run on the current Monsoon - this is my u-bo418. The atmosphere job is currently waiting to run on 114 nodes on Monsoon. I have noticed that this job does not use the GRIB file: the reconfiguration is not switched on.

What this job actually does is governed by the keys on the suite conf → Model configuration page. These select among the various options used by the apps.

I am trying to build a version that runs on ARCHER and this will require finding the various ancillary files, if they exist, and copying them over to ARCHER. For example the model uses eORCA025 GA7.1_AMIP ancillaries which currently do not exist on ARCHER.

This seems to be an experimental set up used only at the Met Office. It may be better to look for a more standard model to base your work on.

Willie

comment:6 Changed 6 weeks ago by willie

Hi Markus,

I now have an ARCHER version which compiles. This is u-bo443. It is currently failing in atmos_main because the easyaerosol netCDF files are not yet present on ARCHER. These amount to 1.8TB and will take at least 35 hours to transfer from the Met Office.

I had to switch to a 19820101 start dump to match the ancillary files used in the model. I don't know what the other start dumps referred to in the model are for.

This model also needs the data in the hadom and hadmn directories under /work/n02/n02wmcginty on ARCHER to work. These are user names of people at the Met Office.

Willie

comment:7 Changed 6 weeks ago by willie

PS,

In your roses directory you need to check out the branch,

fcm co fcm:um-br/dev/nicolasbellouin/vn10.6_easyaerosol_v2

as the model uses metadata in that branch.

Willie

comment:8 Changed 6 weeks ago by willie

Hi Markus,

The easyaerosol data has been transferred to ARCHER in /work/n02/n02/wmcginty. However, the suite u-bo443is currently failing in the UM (atmos-main task) due to a segmentation fault in the first time step. These faults can be very time consuming to debug. It may be worth looking for an alternative suite for your work.

Willie

comment:9 Changed 5 weeks ago by mtodt

Hi Willie

Sorry for the delayed response, thanks a lot for your efforts! I chose that suite for comparability with earlier simulations, but I'll see whether I can find an alternative.

Cheers
Markus

comment:10 Changed 5 weeks ago by mtodt

Hi Willie

I've checked out and am trying to run the suite you created (as u-bo757), but the atmos-main task doesn't submit and is repeatedly (re-)trying to do so. What could prevent it from submitting? I understand it that your suite failed during that task but at least could submit it.

Also, would it be possible to get the easy_aerosol files from n96e/ on ARCHER as well?

Many thanks in advance!

Cheers
Markus

comment:11 Changed 5 weeks ago by grenville

Markus

you are trying to run on 150 nodes in the short queue - the limit is 8 nodes in the short queue.

#PBS -A n02-FEBBRAIO
#PBS -q short
#PBS -l select=150
#PBS -l walltime=00:15:00

see /home/mtodt/cylc-run/u-bo757/log/job/19820101T0000Z/atmos_main/10/job-activity.log — the error message indicates a problem with the job submission.

Are these files not what you need:
/work/y07/y07/umshared/CMIP6_ANCIL/data/ancils/n96e/timeseries_2014-2100/VolcanicAod/v3
/work/y07/y07/umshared/CMIP6_ANCIL/data/ancils/n96e/timeseries_1850-2014/VolcanicAod/v3

Grenville

comment:12 Changed 5 weeks ago by mtodt

Hi Grenville

Thanks!

I don't know much about the easy_aerosol model, but I don't think I can use those files. The following files are requested by the suite:

easyaerosol_files(1)='easy_extinction_sw_PI-MACv2-SP-CMIP6-Volc_1949-2015_v2.nc'
easyaerosol_files(2)='easy_absorption_sw_PI-MACv2-SP-CMIP6-Volc_1949-2015_v2.nc'
easyaerosol_files(3)='easy_asymmetry_sw_PI-MACv2-SP-CMIP6-Volc_1949-2015_v2.nc'
easyaerosol_files(4)='easy_extinction_lw_PI-MACv2-SP-CMIP6-Volc_1949-2015.nc'
easyaerosol_files(5)='easy_absorption_lw_PI-MACv2-SP-CMIP6-Volc_1949-2015.nc'
easyaerosol_files(6)='easy_asymmetry_lw_PI-MACv2-SP-CMIP6-Volc_1949-2015.nc'
easyaerosol_files(7)='easy_cdnc_PI-MACv2-SP_1949-2015_v2.nc'

When comparing files in the directory you pointed out with the ones Willie copied over from MONSOON (granted that's for N216), they differ in size by a factor of ~400. So I assume the volc_aer_… files aren't sufficient.

Cheers
Markus

comment:13 Changed 5 weeks ago by grenville

Markus

I don't know how big they should be, those are the files sync'd for CMIP; they're the files from the MO.

I don't follow what you are trying to do.

Grenville

comment:14 Changed 5 weeks ago by mtodt

As I said, I don't know what easy_aerosol is doing or what the particular netCDF files are needed for. But the easy_aerosol files Willie copied over to ARCHER, which are used by the N216 suite, amount to 1.8TB. The files in

.../umshared/CMIP6_ANCIL/data/ancils/n216e/timeseries_1850-2014/VolcanicAod/v3

amount to 9.2GB and the files in

.../umshared/CMIP6_ANCIL/data/ancils/n96e/timeseries_1850-2014/VolcanicAod/v3

amount to 4.1GB. So I assume that the files in those latter directories do not contain all data that's needed for easy_aerosol.

comment:15 Changed 11 days ago by grenville

  • Resolution set to answered
  • Status changed from new to closed

I believe this is no longer relevant

comment:16 Changed 11 days ago by mtodt

As far as I know, there are no easyaerosol files for N96 on ARCHER at the moment. Would it be possible to copy them from /data/d05/hadom/ancil/easyaerosol/1949-2015/n96e on MONSOON to /work/n02/n02/wmcginty/hadom/easyaerosol/1949-2015/ on ARCHER as Willie had done for N216?

comment:17 Changed 10 days ago by ros

  • Resolution answered deleted
  • Status changed from closed to reopened

Hi Markus,

I am in the process of copying these files from the Met Office to /work/n02/n02/ros/hadom/easyaerosol/1949-2015/n96e on ARCHER. I don't have access to Willie's directory so they will be in my equivalent one. I'll let you know when the transfer is complete.

Cheers,
Ros.

comment:18 Changed 10 days ago by mtodt

Hi Ros, thanks a lot!

Cheers
Markus

comment:19 Changed 9 days ago by ros

Hi Markus,

The files have now been copied over to ARCHER. I did manage to get them put into Willie's directory so they can be found at: /work/n02/n02/wmcginty/hadom/easyaerosol/1949-2015/n96e.

If you will be requiring these files on the new ARCHER please make sure you plan to move them as we will not be migrating data from Willie's directories.

Regards,
Ros.

comment:20 Changed 9 days ago by mtodt

Thank you so much!

Cheers
Markus

Note: See TracTickets for help on using tickets.