Errors when adding output variables in ACCESS-rAM3: NetCDF: Index exceeds dimension bound

General
, ,
1

Hello

Has anyone got the following error message before (within directory $reg_$res_GAL9_um_fcst_000)? I got it when I try to add more output variables in ACCESS-rAM3 simulations. I am adding tens of variables so not sure which one cause the problem (I am trying to identify it). Would you have some idea about the possible cause of this error?

???
???!!!???!!!???!!!???!!!???!!! ERROR ???!!!???!!!???!!!???!!!???!!!
? Error code: 40
? Error from routine: NC_PUT_VAR_REAL_1D
? Error message: NetCDF: Index exceeds dimension bound : NF90_PUT_VAR
? Error from processor: 0
? Error number: 93
???

There is also another error message in the job.out file:

[1755496769.314044] [gadi-cpu-clx-0424:250531:0] ucc_context.c:399 UCC ERROR failed to create tl context for nccl

Regards, Qinggang

2

I find the error results from this set of eight variables:

UPWARD SW FLUX ON LEVELS, DOWNWARD SW FLUX ON LEVELS, CLEAR-SKY UPWARD SW FLUX ON LEVELS, CLEAR-SKY DOWNWARD SW FLUX ON LEVELS

UPWARD LW FLUX ON LEVELS, DOWNWARD LW FLUX ON LEVELS, CLEAR-SKY UPWARD LW FLUX ON LEVELS, CLEAR-SKY DOWNWARD LW FLUX ON LEVELS

3

Hi @qinggangg, it might be that you need to extend the number of records allowed into your output file.

Have you made a new output stream?

If so, my first piece of advice would be to look at the settings for the “_pb” stream in app/um/opt/rose-app-stashpack6.conf and copy the reinit settings from there into your stream.

If you did not create a new stream, please come back and I will ask a few more questions.

5

Hi @cbengel, thank you. I can output more other variables, but only these eight cause the problem.

I do not make a new output stream.

In addition, I tried to add more variables from COSP satellite simulators (see COSP), but all values are zero from all the COSP variables. Does ACCESS-rAM3 support COSP simulators now?

6

Hi @qinggangg,

I wonder if the problem is related to the netcdf output. Are you open to working with FieldsFile output?

If so, you could switch to the u-dg768/access_rel_1_ff branch.

Is that something you would be open to trying?

In regards to COSP, I haven’t heard of anyone here using the COSP simulators. I will have to get back to you on whether that is possible with the u-dg768 suite.

7

Hi @cbengel Thank you.
I think I am fine without these variables for now. If I would like it later, I may try FieldsFile.
I would love to hear the update about COSP with ACCESS-rAM3.

1 Like
8

I have had this error when I try to write a X by Y variable into Y by X space.
It can also happen trying to write a U (cell edge) grid into a T (cell centered) grid.

Given the number of permutations to the possible error, it is unlikely that we will be able to guess the correct answer for what the problem is. I offer this as a potential mental jog for you to find the right direction in your solution.

1 Like
9

@cbengel running the COSP simulators is definitely something the Centre wants the capability to do. We can take the discussion off this thread but I am keen to implement this if it isn’t already.

2 Likes
10

Hi @qinggangg, @griff and @bethanwhite,

There is a mix of issues going on here that we can work on offline.

I will close the topic for now, but will report back once when we find a working solution (to COSP and the other variables).

3 Likes