Opened 8 years ago

Closed 8 years ago

#815 closed help (fixed)

Restart a run

Reported by: SimonDriscoll Owned by: willie
Component: UM Model Keywords:
Cc: Platform:
UM Version: 6.6.3



I was just wondering what I need to do to restart a run from the start dumps. So my run begins at 2080 (a control run), and I've changed numerous relevant dates in that to correspond to a run starting at 2080. However, it (xgyte) crashed yesterday. It's a copy of a job that has run out to the full time (20 years), but crashed around five. I don't know why.

Error messages aren't similar to the ones I've seen before:

Rank 25 [Tue Mar 13 21:04:28 2012] [c7-1c1s6n1] application called MPI_Abort(MPI_COMM_WORLD, 9) - process 25

*Rank 12 [Tue Mar 13 21:04:28 2012] [c7-1c1s6n1] application called MPI_Abort(MPI_COMM_WORLD, 9) - process 12

UM ERROR (Model aborting) :
Routine generating error: U_MODEL
Error code: 1
Error message:

ACUMPS: BUFFIN error - see output


I thought I'd try and restart it and see if it runs. Should I run it just by changing the start dumps and then running (and it will 'know' it's five years in, so takes off from there and runs for fifteen) or should I change all the relevant dates etc. to be a run from 2085 (in which case I guess it will run out to 2105???)? (I'm also wondering about model problems if I run it and the dumps/dates aren't set-up right). Or, indeed, should I be trying to run it at all (is there something more serious going on here)?



Change History (2)

comment:1 Changed 8 years ago by willie

  • Owner changed from um_support to willie
  • Status changed from new to accepted

Hi Simon,

Your model runs for 72 time steps and then you get,

Global Net CO2 Flux into ocean (GtC/yr) -2.63844526731435366E-2
Global Net CO2 Flux into ocean - 2nd C NaN

so it looks like the ocean model may be unstable or in error. See #802 where a similar problem occurred (not yet solved).

There are some check setup problems: the user STASH epflux606 contains some items with a broken grid code.

It may be a good idea to increase the level of diagnostics and in the scientific section, section 13, under DIAG_PRN, select "flush buffer if run fails" until you get a solution.



comment:2 Changed 8 years ago by willie

  • Resolution set to fixed
  • Status changed from accepted to closed
Note: See TracTickets for help on using tickets.