To start production on a new machine, at least as things stand now, you're best off using Peter Onyisi's production daemon.  As of this writing (2/21/2002), production is being run on cdf7, cdf10, and cdf22.  If you want to babysit another copy, talk to Peter first if possible.  Here, though, is a summary of how to set it up, for easy reference.

The primary set of daemons is stored in /cdf/data2e/stream_daemons.  There are currently copies in /cdf/data23a/stream_daemons and /cdf/data17b/stream_daemons, controlling production on cdf10 and cdf7, respectively.  If you want to make your own stream_daemons directory, you should probably copy one of the latter, since it won't have Peter's latest/greatest changes, if he's setting any up.  You do not need to keep a stream_daemons directory on the machine you want to run on, though it would be pretty if you did.
You do, however, need a new copy of the scripts in this directory for every machine production will be running on.

First, edit the file run_prod.sh, which is the script that executes the production daemon.  Set the variable PRODCODEVER to the version of production you want to use (e.g. 4.3.0), the variable EXECDIRECT to the directory of your set of stream_daemons (e.g. /cdf/data23a/stream_daemons), the variable PRODUCTIONDIR to the directory you want the daemon to deposit the produced files in (e.g. /cdf/data10a/produced_stream_a), the LOGDIR variable to where you want the logfiles for each production job to be dumped (by default a subdir of $PRODUCTIONDIR, so just set to error_logs) .

Now, you need to link the files you want to have produced (either old production with the raw data banks attached, or raw data) in the directory prod_daemons/prod_spool_hi.  Then type in the following to start up the daemon (on the (this assumes you're sitting in the prod_daemons directory):

tcsh> ./nicer.py ./proddaemon.py

Check the status of the production using:

tcsh> ps axwww

You should see lines resembling:

19269 pts/0    S      0:01 python ./proddaemon.py
 2245 pts/0    SN     0:00 sh ./run_prod.sh /cdf/data23a/stream_daemons/prod_spool_hi/prod_3.18.0int1_123829_2.root
 2572 pts/0    RN   9807:20 ProductionExe /cdf/data23a/stream_daemons/Production/ProductionExe.tcl -i /cdf/data23a/stream_daemons/prod_spool_hi/prod_3.18.0int1_123829_2.root -o /cdf/data10a/produced_stream_a/prod_4.3.0_123829_2.root

This tells you it's working, which is a very good thing.

Some Things to be aware of:

• If jobs are crashing, and the error logs have an error that tells you something about calibration constants  and the variable PROD_PHYSICS_CDF,  you may be trying to use a version of the CDF calibration constants that doesn't work with the run you're processing.  What you need to do is the following:
• Go to the Data Catalog Browser online
• Enter the following in the scrolling list:
• Data Catalog
• Calibration
• CAL: Used Calibration Sets
• Enter data run = (run you're having issues with) and choose to look for PROD_PHYSICS_CDF
• You get out a table that will give you a working version of PROD_PHYSICS_CDF; put this version in the file stream_daemons/Production/setup_managers.tcl (ex uses version 5):
•   ProcessName set PROD_PHYSICS_CDF

  Version set 5
  parmSetName set PROD_PHYSICS_CDF5
• To stop production, create an empty file named prod.stop in the stream_daemons directory.  The daemon will pick up the file when it's finished with the job at hand, and production will be halted until you delete the file.  CHECK FOR THIS FILE IF PRODUCTION DOESN'T SEEM TO BE STARTING.
• The scripts do NOT appear to work under the bash shell; I've found that it's more efficient to just start the daemon under tcsh.  Don't ask me, ask Peter.