Missing field or wrong dump file

Unified Model
1

Hi everyone, I’m running a new SSP2-45 suite on ARCHER2 with UKESM1.2 (same version as used in TerraFirma).

I get the following error 8 years into the run, so it can’t be a missing ancil. I suspect that a file was not written correctly from a previous cycle, but I don’t know how to troubleshoot this. Would be grateful for any advice!

Thank you,
Andrea

???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 1
? Error from routine: READ_LAND_SEA
? Error message: Missing Field or Wrong Dumpfile
? Error from processor: 0
? Error number: 31
???

2

PS suite ID is u-dj147

Also found this, should I just re-start the run in 2022 and hope it fixes this issue?

Land surface model selected is JULES
Conserve UKCA with other tracers? T
FILE_MANAGER: Assigned : /work/n02/n02/adittus/cylc-run/u-dj147/share/data/History_Data/dj147a.da20230101_00
FILE_MANAGER: : Unit : 36 (portio)
IO: Open: /work/n02/n02/adittus/cylc-run/u-dj147/share/data/History_Data/dj147a.da20230101_00 on unit 36
loadHeader: Model Version: 12.1
Surface types present = 101 102 103 201 202 3 301 302 4 401 402 501 502 6 7 8 9 901 902 903 904 905 906 907 908 909 910 926 927 928 929 930 931 932 933 934 935
global_land_field set to 10678
Error in READ_LAND_SEA_MASK: Missing Field of Type 42

3

Andrea

How does this suite differ from one that ran for more than 8 years?

Grenville

4

It looks like there is something wrong with dj147a.da20230101_00 – xconv won’t open it, and uminfo returns “0” for all header data.

So, yes, maybe rerun the last 2022 cycle.

Grenville

5

Thanks Grenville, I’ll try that - would re-triggering the last coupled task of 2022 work or do I need to do a proper restart from the January 2022 restart dump?

6

There isn’t an equivalent run on ARCHER2, only on the Met Office machine

7

I think the answer is yes, re-trigger the coupled task from the Oct '22 cycle - I’m not sure what that means for all the postproc tasks for the Oct cycle, but pp data in /home/n02/n02/adittus/cylc-run/u-dj147/share/cycle/20221001T0000Z looks OK.

It is very weird that the dump is in this state but no error was reported.

Grenville

8

Thanks Grenville, that didn’t work because the dumps are now out of sync:

This is a CRUN
[WARN] The UM restart data does not match the current cycle time
. Cycle time is 20221001
UM restart time is 20230101

I think I will need to do a proper restart from Jan 2022 unfortunately … happy to hear better ideas though!

9

Argh, we’ll need to tweak the history file - hang on.

10

I found this link since (should have looked more thoroughly before posting) - happy to try this on my own and then follow-up tomorrow if it still doesn’t work!

https://code.metoffice.gov.uk/trac/moci/wiki/tips_CRgeneral#RestartingFailingSuites

11

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.