Date: Mon, 25 Jul 1994 10:27:15 -0600
From: Ben Foster <foster@vishnu.hao.ucar.edu>
Message-Id: <199407251627.KAA25557@vishnu.hao.ucar.edu>
To: ccm-users@ncar.ucar.edu
Subject: regen for branch run
Hi again:
Thanks to Linda for helping me with nhtfrq. I am now attempting
a branch run so I can change nhtfrq. I am continuing with a 6-min
time step, and want to branch off the end of my existing run (day 2.0).
This would read regen data from .../oldcase/rest/r0010, and write
.../newcase/hist/h0001.
I tried a 6-hr branch run (NSREST=3, NELAPSE=60), with NHTFRQ=60
(dispose 1 history at end of run), NREVSN='.../oldcase/rest/r0010',
(full path for branch run), and NSVSN='rstrt'. The case name was
changed in the run script.
The model read the regen data from the old case (r0010), ran the
correct number of steps, wrote 1 time sample to h0001, and disposed
it to mss. So far so good. But then it aborted when attempting to
write the new regen data .../newcase/rest/r0001. It died on mswrite
because the local file r0001 did not exist:
SAVDIS:calling mswrite as follows:
mswrite -f TR -nomail -nowait -t 35 -v CTPUBLIC -w btfoster r0001
/FOSTER/ccm2/flx_eqxb/rest/r0001
SAVDIS:Error from mswrite -- UNIX file does not exist
I have noticed in the code that rgnfls is *not* called from restrt
if lbrnch=.true.:
C
C Assign new restart datasets for all continuation runs. Start naming
C with stfnum for branch run.
C
if (lbrnch) then
write (itf,'(i5)') stfnum + 10000
nrfil = 'r'//itf(2:5)
else ...
call rgnfls
Why is rgnfls not called if a branch run?
The writes to regen datasets (wrtrs1 from scan1), and disposal
of regen files to mss (savdis from wrapup) are dependent on
nsvsn not blank (nsvsn.ne.' '). Nsvsn was not blank in
my run, so what appears to have happened is that r0001 was not
assigned, the regen data was written to some unit(s) (nrg?),
and the mswrite failed because r0001 did not exist (because the
assign was not done). Thanks for any help, I'm slowly learning
your system...
--Ben