When using the Tco1279 resolution with a 450-second timestep (as recommended in ECMWF’s standard configuration), the model hangs after about 12 hours of integration, without any error message. It simply stops progressing.
Interestingly, when I reduce the timestep to 300 seconds, the model runs to completion without any issues.
I am wondering what should I do to solve this problem?
Hi @Guokun_Dai,
Thanks for reporting this. I must admit that the highest supported grid resolution (Tco1279L137) is not frequently used by many OpenIFS users, mainly due to the required resources. Therefore we do not have much experience with the configuration requirements for different HPC systems.
We use the 450 second time step on the ECMWF hpc2020 system without problems for runs longer than 12 hours, so we cannot really reproduce this problem here. It would be interesting to hear from other users who have run OpenIFS at this resolution and see if they have had any such issues?
Regards,
Marcus