Amanda-Users

Re: Testing amanda: running amdump "early" and other questions

2003-09-09 11:44:26
Subject: Re: Testing amanda: running amdump "early" and other questions
From: Eric Sproul <esproul AT ntelos DOT net>
To: AMANDA Users <amanda-users AT amanda DOT org>
Date: Tue, 09 Sep 2003 11:39:35 -0400
On Tue, 2003-09-09 at 10:44, Eric Sproul wrote:

> > dumpcycle 2 weeks
> > runspercycle 10
> > tapecycle 20
> > 
> > That seems right, but it's a little confusing (too much Googling can be 
> > a bad thing :)
> > 
> 
> Yes, that looks correct.  You are running 10 times within your 2-week
> cycle (weekdays only), and you have two dumpcycles worth of tapes in
> rotation, so you don't risk overwriting your only full dump of
> something.

I should proofread better before posting....  After reading Paul's post,
I realized that I was wrong.  If you have 20 tapes, you should set
dumpcycle "1 week" and runspercycle "5".  That way you'll have two
dumpcycles worth of data, and AMANDA will try her darndest to make sure
you have a full dump of every DLE each week.

Setting a 2-week dumpcycle isn't wrong, it's just going to mean that you
may have more incrementals in between full dumps, increasing the work
required to fully restore a given DLE at any given time.  You'll also
need twice as many tapes in order to maintain the 2-dumpcycle paradigm,
so you'd need 40 tapes.  However, it may be necessary to extend the
dumpcycle if AMANDA is not able to fit all the level-0's in 5 tapes.  It
usually takes AMANDA at least one full dumpcycle for the planner to
settle out into a routine, since part of its calculation involves
analyzing past runs.

HTH,
Eric

-- 
Eric Sproul
nTelos OSS Engineering
----------------------
On US Rocket Launcher - "Aim towards Enemy"