Dear Ming,
Thanks for your reply! I have deleted the settings of sst update, but the problem remains unresolved.
I tried two-nested domains, wrf.exe works well. Compared with three-nested settings, I haven't changed anything except removing the outermost nesting. File as attached.
Unfortunately, for research purposes, I have to use three-nested domains. With three-nested, I tried running different combination of nodes and ntasks, I tried changing the driving data from ERA5 to FNL,I also tried to shrink the size of d02 and d03 to the minimum needed. But the wrf.exe crashed every time.
What I find interesting is the rsl.error.0000 always end with 6th step, such as:
1:
Timing for main: time 2018-08-21_00:00:06 on domain 3: 3.49978 elapsed seconds
Timing for main: time 2018-08-21_00:00:12 on domain 3: 0.27070 elapsed seconds
Timing for main: time 2018-08-21_00:00:18 on domain 3: 0.27109 elapsed seconds
Timing for main: time 2018-08-21_00:00:24 on domain 3: 0.27109 elapsed seconds
Timing for main: time 2018-08-21_00:00:30 on domain 3: 0.27150 elapsed seconds
Timing for main: time 2018-08-21_00:00:30 on domain 2: 9.26382 elapsed seconds
2:
Timing for main: time 2016-10-06_00:00:04 on domain 3: 2.92672 elapsed seconds
Timing for main: time 2016-10-06_00:00:08 on domain 3: 0.17259 elapsed seconds
Timing for main: time 2016-10-06_00:00:12 on domain 3: 0.17245 elapsed seconds
Timing for main: time 2016-10-06_00:00:16 on domain 3: 0.17269 elapsed seconds
Timing for main: time 2016-10-06_00:00:20 on domain 3: 0.17290 elapsed seconds
Timing for main: time 2016-10-06_00:00:20 on domain 2: 7.59697 elapsed seconds
There might be some clues? Or there is a problem with my domain settings?