Allow short term archiving to work with MPAS files

[rljacob]

Allow the case.st_archive script to work with mpaso and mpascice history and restart files.

Also should work with mpasli but not tested.

From the case directory, executing ./case.st_archive should move all history and restart files to the short term archive for all ACME components.

Fixes #1305
S2-131
[BFB]

[rljacob]

@jonbob do MPAS land ice model output files start with "mpasli"? I could add that.

[jonbob]

yes, they should

[rljacob]

@matthewhoffman do mpas-land ice history files start with "mpasli" ?

[matthewhoffman]

Yes they do.

[rljacob]

Further testing reveals that this works for STOP_N = REST_N. In the case where interim restart files are made (REST_N < STOP_N), this will delete the interim AND final MPAS restart files. More development needed.

[rljacob]

Pointing @sarich

[rljacob]

Better summary of current problem is here: ESMCI/cime#1364

[rljacob]

@sarich, I've been testing this branch in this case dir on blues:
/lcrc/group/earthscience/jacob/ACME/cime/scripts/Wlowrmaster
You can see the archive results in
/lcrc/group/acme/jacob/archive/Wlowrmaster

To set up your own test, build an low res fully coupled case:
./create_newcase -case NAME -compset A_WCYCL1850 -res ne11_oQU240

On anvil, I'm running this with 96 tasks:
./xmlchange NTASKS=96
./xmlchange ESP_NTASKS=1

I was running for 2 years but 4 months should work:
./xmlchange HIST_OPTION=nmonths
./xmlchange STOP_OPTION=nmonths
./xmlchange REST_OPTION=nmonths
./xmlchange STOP_N=4
./xmlchange REST_N=4
./xmlchange HIST_N=4
./xmlchange DOUT_S=TRUE

The above should run for 4 months, output 1 restart set at the end and automatically call the short-term archiver and everything should be moved to the archive correctly.

Now change REST_N:
./xmlchange REST_N=2
and rerun (delete all the old files in the rundir and the archive first or its easy to confuse things).
This should run for 4 months and output 2 sets of restarts. But the short term archiver will delete all the mpas restarts instead of just the ones from the middle of the run.

You can leave DOUT_S as false and invoke the short-term archiver manually from the caseroot.

[rljacob]

@sarich did my suggested test with nmonths work for you? I'm getting errors about nmonths being an invalid opt_n. I'd hate to go back to doing years.

[sarich]

I got an error too (maybe the same one) so I gave up on it. I'll take another look at it to see if it's something simple to fix.

…

On Fri, Apr 28, 2017 at 4:26 PM, Robert Jacob ***@***.***> wrote: @sarich <https://github.com/sarich> did my suggested test with nmonths work for you? I'm getting errors about nmonths being an invalid opt_n. I'd hate to go back to doing years. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1334 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEs_un8fAwBmnpmNprMSkbutfhzP5hJlks5r0lmTgaJpZM4MkTs0> .

[rljacob]

I did a test with a 2 year run and restarts every year. I also tried RESUBMIT and everything worked!

[sarich]

Too late to be relevant for this issue, but for the future --

@sarich did my suggested test with nmonths work for you? I'm getting errors about nmonths being an invalid opt_n. I'd hate to go back to doing years.

./xmlchange HIST_OPTION=nmonths
./xmlchange STOP_OPTION=nmonths
./xmlchange REST_OPTION=nmonths
./xmlchange STOP_N=4
./xmlchange REST_N=4
./xmlchange DOUT_S=TRUE

You also need to set HIST_N

From src/drivers/mct/shr/seq_timemgr_mod.F90:
Line 771:

 call seq_timemgr_alarmInit(SyncClock%ECP(n)%EClock, &
          EAlarm  = SyncClock%EAlarm(n,seq_timemgr_nalarm_history),  &
          option  = history_option,      &
          opt_n   = history_n,           &
          opt_ymd = history_ymd,         &
          RefTime = StartTime,           &
          alarmname = trim(seq_timemgr_alarm_history))

and at Line 1397:

case (seq_timemgr_optNMonths)
       call ESMF_TimeIntervalSet(AlarmInterval, mm=1, rc=rc)
       if (.not.present(opt_n)) call shr_sys_abort(subname//trim(option)//' requires opt_n')
       if (opt_n <= 0)  call shr_sys_abort(subname//trim(option)//' invalid opt_n')