Hi -

   A follow up to my earlier question - as Joop Verdoes pointed out -
if you set AUTOQUERY to YES and start butc with the -noautoquery flag,
the backup system will run your mount and unmount scripts.  Thanks
Joop!

  If anyone is interested is some of the issues that I came across -
I'll summarize here.  If you are interested in the scripts, please
feel free to drop me a note.  They are pretty simple (all I'm capable
of writing).

  I have a Digital TL820 jukebox which holds 264 tapes.  The tapes in
the box have barcodes on them.  The software intended to be used is an
augmentation of Legato's Networker.  Some of the tapes in the jukebox
are being used as Networker tapes and some as AFS tapes.  I first used
the Networker product to label all of the tapes in the box - telling
it to match the label to the barcode on the outside of the tape, and
to divide the tapes into various "pools" - one pool being used just
for AFS.  Networker is then told not to use AFS pool tapes for itself.

  I then relabeled the AFS pool tapes using the backup labeltape -
giving each tape a -pname that matches the barcode. (This isn't
really necessary, but helps to keep track of things).

  The two main hurdles to overcome are automating the process of
finding and loading a new (not been written to be AFS) tape, and
translating AFS tape names into barcode names.

  To solve the first - I create a file which contains the barcode
names of all of the unused, but available to AFS tapes.  

  To solve the second - when the stCmd_NextTape script is called (with
the dump, appenddump, savedb, or restoredb operation request), it
looks in a database file to determine if it needs a new tape.  Each
line of the database file just contains the AFS tape name, and the bar
code label new for a tape.  If the AFS tape name is in the database,
the script extracts the barcode label and requests that the jukebox
load that tape.  If the AFS tape name is not in the database, the first
tape from the available tape file is used (and that tape name removed
from the file), and the requested AFS tape name and the barcode label
name are written to the database file.  It then requests the jukebox
to load the new tape.

  The load command is:
  /usr/bin/nsrjb -l -nv -f $devicefile "$filename                          "
  (there is a bug in the nsr code such that all tapes have labels 32
  characters long [$filename - the barcode label - is 6 characters])

  The unload command is:
  /usr/bin/nsrjb -u -v -f $devicefile
  
  If the operation is a restore - the tape is requested by pname (bar
code label name).  If you don't pname the tapes, you can just look up
the barcode name in the database file.

  So far, so good.  What I haven't yet tested is crashing the jukebox
and asking the jukebox and the Networker product to reinventory
things.  You are supposed to be able to do this just reading barcode
labels (rather then reading tape labels), but I haven't tested it yet.

  Also haven't tested using several of the tape drives at the same
time (different butc processes).  The jukebox software is supposed to
be able to determine if the robot is in use and wait until it is
available so I don't think it will be a problem.

thanks for the input and listening to this dribble.....
-dave

Reply via email to