Error with layout of masks in MOM6

COSIMA
1

I am setting up another run with panan-005 where I coarsened the bathymetry such that each 2x2 gridbox has the same depth. I followed all of these steps and have the files I need but get the following error during initialisation:

FATAL from PE 0: fms_io(parse_mask_table_2d): mpp_npes() .NE. layout(1)*layout(2) - nmask for Atmos model

Can someone explain to me what this error means?

My work directory is here /scratch/e14/cs6673/mom6/work/panan_005deg_jra55_ryf_bathy_01deg/

1 Like
2

I can’t see e14, but I think this is just saying that the number of cores in your layout doesn’t match the CPUs you asked for. Or perhaps it can’t find your mask table?

3

Thanks @AndyHoggANU . I haven’t changed the layout compared to the control panan-005

`LAYOUT = 144,36

IO_LAYOUT = 36,9`

and still request 3916 cpus.

But I did notice that the name of the mask table changed after running check_mask. It was

mask_table.1268.144x36

and is now

mask_table.1260.144x36

But I don’t know what the first value stands for and whether I need to change something.

4

I don’t know what the first value stands for and whether I need to change something.

That’s the number of “masked” domains. If this has changed, it means you need to change the number of cpus used. There are 144x36=5184 domains in your calculation and previously there were 1268 masked domains. That means there were 5184-1268=3916 active domains, which is the number of CPUs that were requested.
Now you have 1260 masked domains, so you need 5184-1260=3924 CPUs instead.

So it should be enough to update the number of requested CPUs in the payu config to get this running correctly (assuming you’ve updated the mask filename where needed).

2 Likes
5

Thanks @micael, you were right and it started running now and made it to day 3 :raised_hands:

That explanation was really useful, I tried multiplying all sorts of numbers to figure out what that first number might be…

1 Like