search

unifhy-org / unifhy

A Unified Framework for Hydrology
https://unifhy-org.github.io/unifhy
BSD 3-Clause "New" or "Revised" License
11 stars 5 forks source link

Explore potential memory leak(s) #56

Open ThibHlln opened 2 years ago

ThibHlln commented 2 years ago

When running cm4twc on JASMIN for a long simulation period (12 years) on a small region, the run crashes after ~7 simulated years because it runs out of memory. After restarting the run where it stopped, it runs to completion the second time around.

For a period of (1x2+10) years, the temporal resolutions of the components are:

For a region spanning (56.0, 57.5) degN, (-5.5, -2.5) degE, the spatial resolutions of the components are:

SLURM configuration used on JASMIN were:

#SBATCH --partition=short-serial
#SBATCH --exclusive
#SBATCH --mem-per-cpu=16000
#SBATCH -o %j.out 
#SBATCH -e %j.err
#SBATCH --time=24:00:00

The fixed memory leak related to ESMF use in cf-python certainly eliminated the main memory leak in cm4twc, but it seems that there is/are (an)other memory leak(s) still present.