On Wednesday 02 July 2003 23:18, Michael D. Schleif wrote:
>Thank you, all of you, for your help -- I have a working amanda!
>
>Now, I am trying to understand amanda's built-in backup strategy.
> How can I know when `disk' will be backed up full or incremental?
>
>I do not have a changer, and I have more data across all disks than
>single tape capacity. Therefore, I want to stagger full backups,
> such that on any given tape, only one (1) system is full backup and
> all others are incremental backups.
>
>I have scoured amanda(8), and learned alot; but, I do not find this
>information there. Please, kindly point me to the right point in
> TFM, and I can read . . .
>
>What do you think?
Amanda tries to adjust the schedule to make use of a consistent amount
of tape per nightly run.
Example: runtapes 1, dumpcycle 7 days, runspercycle 7, tapecycle 28
2 machines, totalling around 70Gb between them, but the second machine
isn't getting 100% coverage, so a full of everything is maybe 50Gb.
Tapes are DDS2, 4Gb without compression, actually somewhat less in the
real world.
All disklist entries started out using 'best' compression, but the
emails will quickly tell you which ones to skip the compression of
as they'll grow, such as a dir full of rpm's or tar.gz & tar.bz2's.
Starting out by commenting enough entries out of the disklist for the
first run so that a full of everything will pretty well fill one
tape. For the next nights run, uncomment about that much more of the
disklist, and continue in this manner until its all exposed.
Then in keeping with useing 'bumpdays' as a trigger, amanda will start
to advance the level 1's to level 2's etc etc, always with a target
of useing the same amount of tape every night. With bumpdays at 2, I
found it was making amanda constantly jiggle things around when they
were already pretty optimum, so I raised that to 3 days now. But
thats after many dumpcycles to stabilize, so I am now filling a DDS2
tape to 90% or better every night, and the dumpcycle means everything
gets a level 0 in at most a 7 day period.
Amanda is not, and never was, designed to do fulls on friday night,
and incrementals the rest of the week. Either way one will have to
backtrack 7 tapes in the useage schedule and recover in sequence to
have the latest snapshot if the failure is on a friday. To us, its
the same end result, and much more efficient useage of the media if
done amanda's way.
If you find that once in some semblance of balance, that amanda is
using less than 75% of a tape every night, then it might be safe to
consider reducing the dumpcycle and runspercycle to 1 less, then give
amanda a dumpcycle or 2 to settle, look at the useage in the emails
and adjust some more. Conversely, if a level 0 has to be put off by
a day, then it might be indicating you should increase the dumpcycle
and runspercycle to 8 or 9 days. But that would be an indicator too
that you really do need a larger capacity tape drive, or that you
need to increase runtapes. If you do this, you will need a changer,
or sufficient holding disk so that the second (and subsequent) tape
can be 'amflush'ed in the morning when someone is there to change the
tape.
This is somewhat simplified of course, but I've tried to hit on the
highlights of things to consider.
tapecycle seems to be the most miss-understood value, particularly the
reason it should be at least 2*runspercycle*runtapes.
The reasoning behind that is so that one will never be overwriting the
only full of that disklist entry with a new full, but for some reason
it fails, and you therefore have no full at all.
HTH
--
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz 512M
99.26% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2003 by Maurice Eugene Heskett, all rights reserved.
|
