Amanda-Users

Re: Defining a sensible backup routine

2006-04-17 13:39:18
Subject: Re: Defining a sensible backup routine
From: Jon LaBadie <jon AT jgcomp DOT com>
To: amanda-users AT amanda DOT org
Date: Mon, 17 Apr 2006 13:34:38 -0400
On Mon, Apr 17, 2006 at 05:00:25PM +0100, Anne Wilson wrote:
> On Monday 17 April 2006 16:47, Ross Vandegrift wrote:
> 
> > Another thing you can do that helped me out a lot - change the bump
> > parameters!  I work with a fairly large Amanda setup where disk space
> > is usually at a premium.  I was able to save a lot of space by tuning
> > the bump parameters to do more higher-level incremental backups.
> >
> > While the system previously wasted a lot of diskspace in the second
> > half of the cycle doing level 1 and 2 backups, now it usually bumps
> > every day, hitting level 5 or 6 by the end of the week.
> >
> > Of course you take a hit on restore time, but restores are pretty
> > decent from virtual tape as it is, so I'm not too concerned about
> > there being five or six steps instead of two or three.
> 
> Hmm, now we have to see whether I've understood this correctly.
> The situation, IIUI, is that if a backup is close to being a
> full backup it will bump it right to level 0, but also bumping
> is used to balance out the workload in some way.
> No, I don't understand it.

promoting == moving to a lower numeric level
bumping   == moving to a higher numeric level

If the last dump of a DLE was an incremental, and it is not
scheduled for a level 0, then amanda must decide to promote,
repeat, or bump the level.  This decision is based on the
estimated sizes of each of the three levels and is influenced
by parameters bump{size|mult|days} (by default 10MB, 1.5/level,
and 2 days respectively).

For Ross' vtape based backups size was a consideration and he
wanted rapid, easier promotion (lower all values).  For my ptape
based backups that was not a problem and I seldom see anything
higher than level 2.

Independent of the planning for "todays" dumps, amanda also
considers the longer range schedule.  If the total size of
the estimated dumps in the above planning is small relative
to what amanda considers "balanced" (Size of all level 0s
divided by their respective runspercycle) then amanda may
promote some DLEs all the way to level 0 hoping to reduce
the oversized dumps anticipated in the future.  AFAIK only
one parameter influences this, maxpromotedays, default 10000.
It is the maximum number of DLEs that can be promoted for
long range planning.  Set it to zero and nothing will be
promoted for this purpose.  A setting of 1 or 2 may reduce
promoting drastically at some sites.

> So - would you mind being quite specific in what I would need
> to do?  It sounds as though your method would match my needs.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)