[MITgcm-devel] mnc error message

[Martin Losch]

Hi Ed,

unpredicably I get a strange problem on our Cray XD (so far only on  
that machine). In the present case (and this is typical), the run  
terminates after many timesteps (138 days of model time with  
deltaT=600sec = 19872 timestep) and many (138 in this case)  
successful dumps abruptly with these error messages (for each CPU):
> xd1-420-6::run00> cat STDERR.0029
> (PID.TID 0029.0001) *** ERROR *** MNC ERROR: variable 'XC_max' is  
> already defined in file 'monitor_grid.0000000000.t016.nc' but using  
> a different grid shape
And then of course:
> ABNORMAL END: S/R MNC_VAR_INIT_ANY

The monitor_grid.* files are not touched (have the date and time of  
the beginning of the run).
I will try to reproduce this problem, but in the past I failed to do so.
Now, I am not saying that our Cray is configured perfectly, but what  
can cause the model to try again writing the monitor_grid.* files? It  
is a complete miracle to me.

(I am using a domain of 360x200x37 grid points on 32CPUs a 45x50x37  
points, overlap is 3)

Any idea?

Martin