CMA only needs to write information into the database. The CMA
writing process is probably best run on the site server.
The database czar starts up the CMA write process by typing
/$SNODB_VER/scripts/snodb.perl -cma_write.
This process then runs continuously, looking for new files in standard
``titles'' format named *.dat to appear in the directory defined
in scripts/snodb_lib/init.pl. The default is /write_to_database/cma/todo. (Note that this
directory structure is loosely modelled on the standard SNODB/HEPDB
structure.) The files in the todo subdirectory are
processed in time-ordered FIFO sequence by the
/$SNODB_VER/exes/sdb_input_titles program which reads in
the files and submits the associated updates to the database.
To stop the CMA write process, type snodb.perl -cma_stop which creates a file named signal.stop in the appropriate directory and eventually causes the CMA write process to stop. The signal.stop file is then deleted.