ANCIL_ANTS failing

Unified Model
, ,
1

I’m trying to run a copy of my own UM job that ran OK a couple of months ago. The suite id is u-cx568, and it is vn11.9 nesting suite.

I start with what I have done so far and add some notes for future reference. My question is at the end.

ANCIL_ANTS_VEGERAC is failing and the first problem appeared to be that qrparm.mask_cci was not created for some reason. So I copied it over from a suite run previously.

So now job.err has this;

/bin/sh: /home/d03/myosh/cylc-run/u-cx568/bin/python_env: Permission denied
[FAIL] python_env ancil_lct.py ${source}
[FAIL] --target-grid ${target_grid} --transform-path ${transformpath}
[FAIL] -o ${output_vegfrac} --landseamask-output ${output_lsm}
[FAIL] --ants-config ${ANTS_CONFIG} <<‘STDIN’
[FAIL]
[FAIL] ‘STDIN’ # return-code=126

So it looks like the permission of python_env has been denied? the status was -r–r–r–. so I did chmod +w and then chmod +x on this file.

(But be aware that this is
/home/d03/myosh/cylc-run/u-cx568/bin/python_env
and not
/home/d03/myosh/roses/u-cx568/site/monsoon-cray-xc40/python_env
that required the change.)

Anyway, this did not solve the issue but brought me a different place. Now the error is like this;

OSError: /projects/cascade/hadsw/NS/ANTS_SRCE/vegetation_fraction.ao996.r73298.nc does not exist.

This is the error I have had before and was resolved in my older ticket. That time the file was moved (or the directory was renamed) to cascade.2022-06-07 instead of cascade, and I had to change all lines in all codes containing ‘/projects/cascade’. doing ‘grep -r cascade’ will find all occurrences.

Now however, either /projects/cascade, /projects/cascade.2022-06-07, or anything similar does not seem to exist.

So here is my question. Where is vegetation_fraction.ao996.r73298.nc now?

I hope that if someone moves files or directories, they create symbolic links in the original location.

Thanks for your help.
Masaru

2

Hi Masaru,

On Monsoon2 when a project (or user account) has finished/expired they are eventually earmarked for deletion. Before being permanently deleted they are moved to a <dir>.year-month-day for a period of time.

The Cascade project has now been deleted.
I can only suggest talking to Stu Webster to see if he might have that file somewhere else.

Regards,
Ros.

3

Permanently deleted? How the hell is it possible? What happens to people whose jobs rely on those data??? Of course I don’t mean to blame you personally, but that is absolutely insane, isn’t it?

But there must be an alternative, right? Or otherwise how do we create ancillary files?

Masaru

4

Hi Masaru,

There is only a finite amount of disk available, space has to be reclaimed from finished projects eventually, otherwise there would be no space for new ones. The project management process is detailed in the Monsoon2 documentation; the Monsoon2 team work with the Project/Tech PI to close expired projects and remove them from the system. The moving of the project data directory for a period (usually 6-12months) before final deletion, is done to try and alert those who are still using it and give them time to move required data to another location. It is impossible to identify if someone is still using a file.

I’ve had a quick look on the MO internal HPC but directory structure is different. Sorry, but I can only suggest contacting Stu Webster, who owned these files to see if he has it stored somewhere else. If these were standard nesting suite files they should still be available somewhere.

Best Regards,
Ros.