HOW TO RUN PRODUCTION AT SLAC
                         -----------------------------

This file is at:

$BFROOT/doc/Computing/Simulation/prod/howtorun.txt

and can be seen at URL:

http://www.slac.stanford.edu/BFROOT/doc/Computing/Simulation/prod/howtorun.txt


                                 INSTRUCTIONS
                                 ------------

Note that the ~/prod directory is now obsolete.

In all cases
------------
o login to a morgan machine


Only at major updates of production runs (only by site responsible)
-------------------------------------------------------------------
o cd $BFROOT/prod/SimProd

o mv defineRunsxxxx defineRunsxxxx_old

o cvs update -A 
or, if there are new files in subdirectories (to be notified by prod
manager):
cvs update -d 
followed by 
cvs update -A

o Check that new defineRunsxxxx looks ok. If all ok 
rm defineRunsxxxx_old

o cd $BFROOT/prod/mgr

o edit defineRunsxxxx
and uncomment those runs which are assigned to your institute

o defineRunsxxxx
where eg xxxx = 9703 for March '97 and 9704 for April '97 etc.


To Run a Given Job at SLAC (number nnnnnn)
------------------------------------------

o $BFROOT/prod/mgr/runjobs.pl nnnnnn
(you can give more than one job number, separated by spaces eg. 
runjobs.pl nnnnn1 nnnnn2 nnnnn3
etc.) The script checks that the runs have been defined, and that they
have not already run. The script will stop you trying to run it from
machines other than morgans.

o To check progress, type 
bjobs or bjobs mmmm 
where mmmm is the job number
Note that you can check on the whole batch system with 
bjobs -u all
or on the jobs of any user with
bjobs -u user
where "user" is their id.

o Job log will be temporarily stored in 
/nfs/juno/u52/prod/log/out/nnnnnn
You can watch it grow with ls -l /nfs/juno/u52/prod/log/out/nnnnnn

o To find the job output, type:
ls -l /nfs/juno/u52/prod/log/out/nnnnnn
or
ls -l $BFROOT/prod/log/allruns/nnnnnn
The .xdr file is in each case a symbolic link to a file in the staging area.
See its size by typing: 
ls -l path 
where "path" is the path indicated by the link displayed using the ls -l command above.

o When job ends:
Check that correct number of events has been generated by looking at 
end of $BFROOT/prod/log/allruns/nnnnnn/runnnnnnn.log
Also check the job log for important warnings and errors.

o Check that xdrCatnnnnnn.out looks ok

o Update database by typing:

$BFROOT/prod/mgr/dbupdate nnnnnn

and answering the prompts. 

o If all is well, you are ready to copy to silo tape, along with other
runs.


COPYING TO SILO TAPE AT SLAC
----------------------------

o Tape number to write to is now automatically dealt with by the script.

o cd $BFROOT/prod/log/allruns

o For a block of runs, nnnnn1...nnnnnm.
Check that total o/p size is not more than 850 MBytes. Then type eg.:

~raines/prod/dbsilo 043106-043110

It will even take combinations.

~raines/prod/dbsilo 041006 041007 041009-041011

If you give it a '-n' option, it will print out the stagewrt
command and exit immediately doing nothing.

At the end, it tells you exactly what to write on the tape 
log file.

o Edit $BFROOT/doc/Computing/Simulation/prod/silo.html

THAT`S IT!