On Tuesday 21 December 2004 05:49, Shai Ayal wrote:
>Hi all,
>
>I'm trying to set up an backup system in our company. We have two
> major file systems to backup - one is the homedirs of everyone,
> including mail in Maildir format, and the other is data.
>
>Our idea was to have two computers -- one would hold the data and
> homes, and the other would back it up daily to a disk.
> Additionally, once a week or so we would to a tape backup to take
> off site.
>
>The data is quite huge and it is often updated but only in small
>amounts, i.e. we have about 80GB of data, and we add less than 300MB
>each week. Clearly doing a full backup every week is very wasteful.
>
>Homedirs are standard and I think the usual amanda backup algorithm
> will work with no problem.
>
>Problems:
>
>1. Can I implement two different backup strategies, and write the
> result to the same tape? i.e. although the strategy for home and
> data is different, they would go in the same tape if they are
> scheduled for the same day?
I think yes, with different disklist entries. This is basicly
what the disklist is for AND it also allows you to break huge pieces
of data down into manageable pieces.
By managable pieces I mean that for amandas backups to work, 1: no one
backup can be bigger than a tape as amanda will not span one entry
across multiple tapes by design, and 2: multiple tapes can be used by
specifying runtapes as greater than 1, but 1 above still holds.
This works by telling amanda how big say 2 tapes would be, and amanda
will write till it hits the EOT, at which point amanda will change
the tape (if you have a changer and have amanda configured for it)
and then restart, from byte one, the dump file from the disklist entry
that failed on the previous tape.
I am currently using a disklist with 51 entries here at home because
being that fine grained, amanda is free to optimize the schedule and
achieve a 90+% tape usage on each nightly run. No one entry is over
700 megs. This is less than 10% of the tapesize I've set, so once
the schedule gets optimized, it will be able to do 95% or more
tape filling every night. Here, its set to do a full on everything
at some point in a 7 day window.
This level of control can only be achieved by tar since dump
generally covers a complete filesystem. Tar is file oriented making
it possible to break a large main directory into its much more
manageable subdirs.
We've just been informed that tar-1.14 is broken, so make sure the
tar you use is 1.13-19 or 1.13-25.
>2. What is the recommended strategy for backing up the data?
> "strategy nofull" seems appropriate, but does it work? -- the
> manual seems to be ambiguous about it. Does anyone have other
> recommended strategy ?
Amanda cannot do what you are calling a "nofull" without first doing a
full during each cycle, but can then do incrementals based diffs
since that full. It must have a reference full in its database in
order to define whats been changed when doing the incrementals.
Look in the man pages for how the various *cycle stuff works, this is
how you tell amanda the outline of what it is that amanda is to try
and do.
>3. On the same subject, can someone help me with the technical terms
> of backup level: I understand level 0 is full backup, but do level
> 1 backups depend on the last level 0 backup done or on the last
> level 1 backup done ?
AIUI, a level 1 depends on diffs from the last level 0, and a level 2
depends on the diffs from both a previous 0 and the last 1 combined.
So a file thats not changed since the last level 1 will not be
resaved in a level 2 run even if it has been changed since the level
0.
>Thanks in advance for any help
>
>Shai Ayal
--
Cheers, Gene
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.30% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attorneys please note, additions to this message
by Gene Heskett are:
Copyright 2004 by Maurice Eugene Heskett, all rights reserved.
|
