770 likes | 1.13k Views
Introduction to ACCESS-UM IT infrastructure. A.Sulaiman Martin Dix Michael Naughton Stuart Moore Acknowledgements: NEC Supercomputing Australia Glenn Greed (UKMO) Ilia Bermous, Peter Steinle, Xi Xiao, Tan Le and others who have been working on the ACCESS project. AGENDA.
E N D
Introduction toACCESS-UM IT infrastructure A.Sulaiman Martin Dix Michael Naughton Stuart Moore Acknowledgements: NEC Supercomputing Australia Glenn Greed (UKMO) Ilia Bermous, Peter Steinle, Xi Xiao, Tan Le and others who have been working on the ACCESS project.
AGENDA • Introduction • Introduction to UM • UMUI • Computing Environment overview • Introduction to UMUI • Brief Introduction to TRAC -- Break -- • FCM & UM 6.3 Development
UMUI Unified Model User Interface
UMUI Introduction Unified Model User Interface • Made up of definition files interpreted byGHUI(Generic Hierarchical User Interface),script and and control file templates and tcl/tk routines.
UMUI Introduction Unified Model User Interface • GUI used for building, setting up and running the UM • Accommodate multiple UM version • Currently vn6.0 and vn6.3 • Client Server (user umui application <-> umui_server) • Currently on d2 only
Running UMUI • Need to set up password-less ssh • Need an account on d2 • Must have full X-Windows connection • Set path to include /access/binx86_64 • copy .umsetvars_vn6.3 from ~access • To start, enter (on the command line): umui&
Opening a UMUI Job • Open xaeza
UMUI Job Submission The following Directories appear (if not already there): $WORKDIR/nwpra/--- Temporary job files $DATADIR/nwpra/--- job output $HOME/umui_jobs/--- job scripts and parameters $HOME/umui_runs/--- run scripts and parameters $HOME/um_output/ --- job output Output name format: $RUNIDxxx.$RUNID.xxxxxx.xxxxxxx.leave eg: nwpra000.nwpra.d07103.t120946.leave YearJulianDay.Time Usegrep ‘End of UM Job’ For quick check if job is successful
Debugging runs Output (ascii) from a UM job appears in $HOME/output/$RUNIDxxx.$RUNID.xxxxxx.xxxxxxx.leave ~ 10,000 lines!!!! There is a learning curve to understand these output files. If the model fails there will usually be a traceback within the output file.
Start Dump Common errors Other than the obvious file does not exist! • “Atmospheric dump contains wrong number of prognostic fields” Number in dump = 776 Number expected = 750 This comes about when the start dump ‘science’ options do not match the UM forecast job science options. Check UM setup and/or use reconfiguration. 2. “Atmospheric basis time mismatch” (start dump time does not match that in UMUI).
Debugging (run-time) If little appears to happen in a run check that the start data and the executable actually exist! • “file not found” or “OPEN” “UNIT” If it fails during timestep 1 • %_ALLOC (not enough memory) • Undefined external (chosen section not compiled) • Atmospheric basis time mismatch (start dump time does not match that in UMUI). If timepstep (>1, < ~20) • Poor surface fields after reconfiguration? Much more tricky to solve, check ancillaries and start dump data for anomalies.
More UMUI Setup Panels • Defining job resources • Physics options • Model Resolution and Domains • Start Dump (Initial Conditions) • Stash (Output fields & files)