I am posting the following for Bill Colwell, who reads this list via our
local Draper newsgroup gateway and cannot post via NEWS because of
the resulting LISTSERV subscriber address mismatch.

 - seb

From: Bill Colwell <BColwell AT draper DOT com>
Organization: The Charles Stark Draper Laboratory
X-Newsreader: IBM NewsReader/2 v1.00
Date: Wed, 27 Oct 1993 21:26:48 GMT

In <1993Oct11.185013.24938 AT draper DOT com>,
 SPGJAF%UCBCMSA.bitnet AT yalevm.ycc.yale DOT edu (Joseph A. Faracchio {510} 
642-7638
 {w}) writes:
>I have thought of a simple solution that others may find is good enough.
>
>Use a 0% low threshold.  This is more palatable with ADSM than it was with
>WDSF because now ADSM caches the info.  All your data is now safely on tape.
>
Depending on how often a user backs up, this may not be a good idea.
Consider someone doing daily backup.  A year from now his/her hard
disk fails.  How many tapes will the server mount to do a full
restore?

>To be sure I'm planning a timer-pop command late at night to change the
>hi to 1 and low with low at 0 it should flush everything onto tape.
>
>I'm planning on using this with a prop routine that watches ADSM and
>copies the tape to a duplicate when it is 'closed full' and maybe once
>a night after I flush the disk stages to tape.
>
>In the end I think there's an ADSM command like WDSF that tells it that
>there is some lost data and then , I assume, it causes someone's next
>incremental backup to really be a 'new' full backup.  Right?
>
A bad tape can be AUDIT'ed.  Inaccessible files will be removed from
the server.  If the file was an inactive version, there will be no new
backup.  If it was an active version, then the file will be backed up
again.  This action does not cause a new full backup - there is no
such thing as a full backup in ADSM.

>Of course this all assumes that sometime in the day/or/night theres a
>'quiet' period that you can count on.
>
>cheers ..joe.f.


Bill Colwell
C. S. Draper Lab
Email: BColwell AT draper DOT com
Voice: 617-258-1550