#759 closed defect (fixed)

Trusting r7040

Reported by: luyssaert Owned by: somebody
Priority: critical Milestone: ORCHIDEE 4.1
Component: Anthropogenic processes Version: trunc
Keywords: Cc:

Description (last modified by luyssaert)

Trusting has been down for a while. It is working again but the latest trusting resulted in a handful of failures:

--> trustO5d: failed
Experiment Type: "OOL_SEC_STO_FG2", compilation: debug, Period lenght:1D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO5d

--> trustO6d: failed
Experiment Type: "OOL_SEC_STO_FG2", compilation: debug, Period lenght:2D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO6d

Hint: the simulations which have been compiled in debug mode don't run. This problem was already reported in ticket #756.

Hint: There is also a problem with XIOS. I think it might be due to a variable in stomate_lpj.f90, maybe rhLitter, because trustO7d (using IOIPSL output) runs but not trustO6d which is the same simulation except for the difference in the IO handling.

--> trustO8d (Using interpolation with XIOS): failed
Experiment Type: "OOL_SEC_STO_FG2", compilation: debug, Period lenght: 2D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO8d

Hint: same problem as in ticket #757.

--> trustO9d (OK_SOIL_CARBON_DISCRETIZATION=y): failed
Experiment Type: "OOL_SEC_STO_FG2", compilation: debug, Period lenght: 2D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO9d

Hint: the consistency test between inconsistency between flux and pool based nbp doesn't pass.

--> trustO10d: failed
Experiment Type: "OOL_SEC_STO_FG4", compilation: debug, Period lenght: 1D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO10d

--> trustO11d: failed
Experiment Type: "OOL_SEC_STO_FG4", compilation: debug, Period lenght: 2D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO11d

--> trustO12d (on regional domain): failed
Experiment Type: "OOL_SEC_STO_FG3nd", compilation: debug, Period lenght: 2D, Total lenght: 4D
/home/scratch01/pcadule/TRUSTING/ORCHIDEE/RUNDIR/trunk_rev7040_20210202T1133/modipsl/config/ORCHIDEE_OL/trustO12d

Hint: put back [Post_1M_stomate_history] and [Post_1M_stomate_ipcc_history] in stomate.card for OOL_SEC_STO_FG2 and OOL_SEC_STO_FG3nd

Other observations from the trusting which do not directly relate to any recent changes. Not sure for the moment whether these are independent problems or just spinoffs from the problems listed above:

  • the calls to the following subroutines fitting_laieff and anthropogenic_mortality in stomate_lpj.f90 generate error messages
  • the calls to the following subroutines fit_laieff_to_points and gauss_jordan_method in stomate_laieff.f90 generate error messages
  • the calls to the following subroutines clearcut_harvest_aboveground_dead_roots, second_coppice_cut, check_mass_balance and thinning_harvest_aboveground_dead_roots in sapiens_kill.f90 generate error messages
  • initialisation of the variables pool_end, check_intern and closure_intern in sapiens_kill.f90 generate error messages, but that might be a false problem

Change History (4)

comment:1 Changed 20 months ago by luyssaert

trustO5d: Fixed in r7057
trustO6d: Fixed in r7057
trustO8d:
trustO9d: Fixed in r7058, r7103
trustO10d: Fixed in r7057
trustO11d: Fixed in r7057
trustO12d: Fixed in r7057, r7103

Last edited 20 months ago by luyssaert (previous) (diff)

comment:2 Changed 20 months ago by luyssaert

  • Description modified (diff)

comment:3 Changed 20 months ago by luyssaert

  • Description modified (diff)

comment:4 Changed 11 months ago by jgipsl

  • Resolution set to fixed
  • Status changed from new to closed

Using revision 7345, all tests in the trusting are running except:

  • trustO8d at obelix. This test is using XIOS_INTERPOLATION=y and is running at irene and jean-zay. See specific ticket #807
  • cross test LO1p=LO4p (mpi_omp = mpi). This test compares a run in hybrid mode with pure mpi mode. It gives the same results at Jean-Zay but not at Irene. If at irene, ORCHIDEE is deactivated, LMDZ is then run in bucket mode, the results are still not the same. Therefore we consider this as an LMDZ issue. It is probably related to the optimization options set in the compilation. These options have been validated by longer runs by the LMDZ and coupled model team and the non convergence is not considered as a problem by them. The trusting at irene is set to OK even if this test is different.
Note: See TracTickets for help on using tickets.