Porting HadGEM3 suite to ARCHER2 (2)

Unified Model
1

Hi Grenville

Thank you very much for assisting us with importing u-dc464 (HadGEM-GC5) to ARCHER2 as u-dj927. I have successfully run it on ARCHER2, and both the post-processing and archiving are functioning well.

Currently, I meet an issue regarding PPTransfer by Globus (the substitute of Gridftp which retired last Friday).

I followed Step 1 and 2 in the instructions (link), and then reload the workflow and started a new run. The pptransfer seems meet a timeout issue with a message in the log file mentioning “Timer is not running”.

In step 2, I additionally add the corresponding IDs of “Archer2 filesystems collection” and "Jasmin default collection” to “Globus_src_coll” and “Globus_dest_coll”, respectively, but met the same issue.

I further checked the connection between JASMIN and ARCHER2, and it seems that I can manually transfer the outputs to JASMIN using the Globus interface.

I suspect the issue stems from the timer setup in my configuration. Do you know how to resolve this?

Many thanks.

Xu
PS. Here are the locations for my copy of u-dj927 (/home/n02/n02/xuang/roses/u-dj927) and the latest run (/home/n02/n02/xuang/cylc-run/u-dj927/run14).

2

Hi Xu

The error is explained in the pptransfer job.err file:

[SUBPROCESS]: Error = 2:
ls: cannot access ‘/work/n02/n02/xuang/archive/u-dj927/19780901T0000Z’: No such file or directory

[FAIL] Error checking files to transfer

This happened because postproc did not have any files to transfer (trapping this is not clear cut, so we don’t). In this case, the workaround is simply to set the state of pptransfer for cycle 19780901T0000Z to succeeded (see Interventions — Cylc 8.3.4 documentation). The suite should then proceed.

Grenville

3

Hi Grenville

thanks for the solution. it works well for the initial tilmestep but afterwards I got following error messages of my PPTransfer for data transfer to JASMIN by Globus.

image
image1920×244 42.5 KB

I further checked the job.err for PPTransfer, it appears that my job cannot use globus command.

13081736782317_.pic_hd
13081736782317_.pic_hd1920×260 47.7 KB

I have checked the connection between ARCHER2 and JASMIN manually by “Globus login” after loading Globus-cli on ARCHER2, which works well. do you have a clue on such bug?

thank you
Xu
ps: the workflow ID is u-dm227 (local copy of your u-dj927). you can find the configurations
here: /home/n02/n02/xuang/roses/u-dm227.

4

Xu

In https://cms.ncas.ac.uk/unified-model/pptransfer-globus/, section 6 it asks that you

Setup a symlink to your .globus directory

ARCHER2> cd /work/n02/n02/<archer2_username>
ARCHER2> ln -s ~/.globus .globus

that appears to not be correct for your account. Please redo this instruction.

Grenville