[H-GEN] Tape Drives

Greg Black gjb at gbch.net
Sun Jan 6 01:23:12 EST 2002


[ Humbug *General* list - semi-serious discussions about Humbug and  ]
[ Unix-related topics.  Please observe the list's charter.           ]
[ Worthwhile understanding: http://www.humbug.org.au/netiquette.html ]

This is an interesting topic and, as must be clear already,
there are at least as many answers as there are people to offer
them.  What this means is that it's important for sysadmins to
develop a mechanism that they can use to select appropriate
backup procedures for their needs and that they need to think
carefully in order to determine just what those needs are.  Here
are a few more thoughts that people might find helpful.

Backups can provide many useful services, including:

  * recovery from fumbling with rm(1) and other user errors

  * recovery from serious bugs in your database software when
    the users bang on it in unexpected ways

  * recovery from the cleaner pulling the power cord out in the
    middle of a major disk write

  * recovery from catastrophic disk failures

  * recovery from theft of equipment

  * recovery from the building burning down

Some of these can only be accomplished if the backup media are
located off-site, either in the form of tapes (or removable
disks for those so inclined) carried away each day or via
network backups to remote sites.  The very process of storing
the backups off-site opens you up to new concerns in the form of
the data security -- is it encrypted or can the bad guys just
read the tape (or wire) with all your confidential customer
data?  And how does your privacy policy deal with this?

Clearly, it's important to be able to recover from disasters,
even if they occur very rarely.  It's also important to be able
to recover from at least certain lesser data loss problems.  The
value of the data and the cost of replicating it are important
considerations; the impact on the organisation in the event that
the data cannot be restored is also critical.

And, once you have determined *what* has to be backed up and
*how frequently* it has to be backed up and *how much volume*
has to be backed up and *what protection* it needs, you have to
select suitable media and suitable software and arrange training
of the operators.

The effort by the operators must be trivial -- the software
should remind them each day to change media; it should tell them
what media to load; it should whine at them if they then manage
to install the wrong tape; it should call the sysadmin if the
operators still haven't got it right; it should monitor how
often the tapes are used and arrange replacement for tapes that
have reached your chosen end-of-life; it should provide proper
encryption and/or compression of the data.

Some people use Amanda to handle at least some of those tasks;
hard-core sysadmins develop their own software because they are
then certain that it meets their own needs.  Of course, if they
don't also practise all the recovery scenarios of interest, they
still have not done much of the job.  And it's essential to
ensure that tapes (if they are the chosen media) can be
successfully and regularly read by at least one alternative tape
drive that will be available in case a restore is needed.

One problem with tape drives is that many drives, despite having
"read-after-write" verification, will in fact fail to read tapes
they have written when presented with those tapes a day or sixty
days later.  Tapes are, nevertheless, most often the cheapest
and most flexible and most transportable media and they are not
going to go away any time soon.

Most sites will benefit from using a range of approaches.  At
home, I rsync all the important stuff around my network so that
I have a range of machines storing the data.  When I go away, I
take my laptop which has its own copy of all that stuff.  I also
run nightly backups (using my own software, of course) to my one
DDS-3 SCSI tape on the Pentium-166 that is my sole SCSI box and
which has a number of housekeeping jobs.  (I also carry enough
recent tapes to restore everything when I'm away from home -- in
case of theft or fire.)

On customer sites, I insist that they have at least two boxes
with identical tape drives and my software arranges for the
tapes to be tested in the "spare" tape drive often enough to be
sure we won't be trapped.  And there too, I rsync the important
data around the network for redundancy.

Like much of this type of software, my backup suite is easy
enough to be used by complete idiots -- because on the customer
sites it is only used by idiots -- but it does require a skilled
sysadmin to set it up in the first place.

Finally, be certain you can answer these two questions:

  * Why am I backing up the particular data?

  * How do I know that my backups are useful in the light of
    question 1?

--
* This is list (humbug) general handled by majordomo at lists.humbug.org.au .
* Postings to this list are only accepted from subscribed addresses of
* lists 'general' or 'general-post'.



More information about the General mailing list