Amanda-Users

Re: dumpcycle, runspercycle questions

2003-08-14 10:07:16
Subject: Re: dumpcycle, runspercycle questions
From: Jon LaBadie <jon AT jgcomp DOT com>
To: amanda-users AT amanda DOT org
Date: Thu, 14 Aug 2003 09:59:12 -0400
On Thu, Aug 14, 2003 at 02:34:14AM -0400, Charles Sprickman wrote:
> Hi,
> 
> Hopefully this is a simple question, as I'm not trying to force amanda to
> do anything unnatural...
> 
> I've been going through the docs and the manpage, and I'm not really
> getting the "big picture".  I've got amanda running, and everything seems
> fine (even have the firewall issues worked out), but I'm still very
> confused as to how amanda decides how often to do a level 0 and how many
> incrementals it will do before doing another level 0.

Its an Amanda 'right of passage' thing.
You are not 'Aman' until you ask these questions. :)

> For example, right now amanda runs every other day.  So I've got the
> following:
> 
> dumpcycle 30
> runspercycle 15
> tapepercycle 24 (this is not a magic number, this is just how many AIT
>                  tapes I can spare)


I'm not sure how you mean "you have amanda running".  It sounds like you
already have it 'in production' and doing regular backups.  If that is
the situation, AND you already have all your DLE's (disklist entries)
active (uncommented in your disklist), then you may not see amanda's
schedule settle down for months and months and months.  Here is what
I mean.

One of amanda's scheduling objectives is to balance the amount of data
taped each run.  To do this it needs a history of prior dumps and if
appropriate, their compressibility.  Any time a new DLE becomes active
amanda must give it a level 0, nothing else is possible.  So, if you
started out with all your DLE's active, on the first amdump you did
a level 0 of everything.  (did you do that?  did it fit on one tape?)

Let's assume you did, and it took 60GB of the 100GB tape (not 60GB
of data, 60GB of tape after compression).  That must be how much
data you have total as I'm assuming all your DLE's were active.
Given that history, amanda will be trying to average 1/15th of that
at a level 0 each amdump run, i.e. about 4GB of tape per run.  More
when incrementals are considered too.  So each amdump should be a
mixture of level 0's and incrementals.

On subsequent runs amanda checks which DLE's 'must' get a level 0,
i.e. 30 days, or 15 runs, since last level 0.  If those DLE's do
not total the desired "4GB of tape" per run, then amanda may elect
to "promote" a DLE or three and do an earlier than required level 0.

This type of consideration is done anew each amdump.  If your disklist
does not change, it may take amanda 2 - 5 times through a dumpcycle
before it gets into a regular rhythm of level 0 dumping.  That is why
I said for your setup, it may be months and months.

Amanda's tape-usage balancing is why little sites like mine and Gene H's
can back up lots of disk capacity on little 4GB or 12GB DAT tapes.

If you were able to fit all your planned DLE's on a single tape on the
very first amdump run, then as my example showed, you will be dumping
very little data on subsequent runs.  You may want to consider changing
your parameters so there is a larger percentage of tape used per run
but more frequent level 0's.  Maybe dumpcycle 8 days, runspercycle 4.
Each DLE will get a level 0 about once a week and your 24 tapes will
hold 6 complete dumpcycles.  BTW really every other day?  or is that
a euphemism for "3 times a week - M,W,F"?

Or you will be able to add lots of extra client hosts and DLE's with
your extra tape capacity.  An Aman gets to choose :)


-- 
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)

<Prev in Thread] Current Thread [Next in Thread>