Install_cold

Hi,
In job u-db898 I’ve turned off reconfiguration with the start file set to /work/n02/n02/tetts/start_files/db898a.astart. I am trying to speed up my simulation. I also cleaned out everything in /work/n02/n02/tetts/cylc-run/u-db898/share/data/History_Data/. But now install_cold is failing. It seems to be failing because it can’t find /work/no2/no2/tetts/start_files/da953a.da19820101_00 . The da953 etc is what it was originally using. What am I doing wrong?

ta
Simon

Simon

u-db898 still has AINITIAL = ARCHER2_AINITIAL_DIR + ‘/da953a.da19820101_00’ (set in site/archer2.rc) - it’s been hard wired at some point.

However, the model starts from astart (AINITIAL is the input to the reconfiguration) - change astart in um->namelist->Model Input and Output->Dumping and Meaning

Grenville

Grenville,
many thanks. @luke I am minded to remove the setting of AINITIAL from my config as that seems to be something that should be explicitly setup when using it!

Hi Simon,

I agree! This was included in the original UKESM1.1 AMIP suite. In fact, there are few things there that I disagree with how the suite was set-up, but I’m not expert enough with how these settings are defined to be sure of how best to change them. Perhaps someone else in CMS might be able to give the suite a once-over and suggest how to improve things.

Many thanks and best wishes,
Luke

Hi Luke, Simon

It’s a cut and paste error from when we were updating all suites for ARCHER2.

The “Set up start dumps” section in the archer2.rc file should be the same as what we had for archer.rc.
Ie:

{# Set up start dumps #}
{% set AINITIAL = ARCHER2_AINITIAL_DIR + '/' + AINITIAL_FILE %}

Then you set the vars to whatever you want in the rose edit GUI.

I’ve added this to the archer2_fixes branch so should get into the trunk at somepoint.

Cheers,
Ros.

Put that into my archer2.rc file.
Simon

I’ve now actually done it. And now install_cold is running out of time with no useful output. Very strange…
@RosalynHatcher any thoughts? [Job is u-db898]
Simon

Hi Simon,

In the job.err file:

slurmstepd: error: *** JOB 5102009 ON dvn01 CANCELLED AT 2023-12-18T09:30:10 DUE TO TIME LIMIT ***

It basically ran out of time. Retrigger it. If it does it again, increase the time.

Cheers,
Ros.

Hi Ros,
thanks a lot. I’ve had this failure before. Re-triggering sometimes works; sometimes not. How do I increase the time limit for cold_install. Does not appear to be something configurable in the rose suite editor. I wonder why it is taking more than a minute. It appears to generate links which are fiarly fast.

Simon

Hi Simon,

If things aren’t available in the GUI then you usually find it in the *.rc files.

See [[INSTALL_RESOURCE]] in the site/archer2.rc file

The /work filesystem is on go slow today. :roll_eyes:

Cheers,
Ros.

Thanks a lot. Yes, I should get in the habit of looking in the .rc files on puma!
And a full filesystem is a slow filesystem…

Simon

This topic was automatically closed 24 hours after the last reply. New replies are no longer allowed.