Amanda-Users

Re: pragmatics of holding-disk space

2005-07-06 19:57:26
Subject: Re: pragmatics of holding-disk space
From: Frank Smith <fsmith AT hoovers DOT com>
To: Michael Hannon <jmh AT physics.ucdavis DOT edu>, amanda-users <amanda-users AT amanda DOT org>
Date: Wed, 06 Jul 2005 18:43:57 -0500
--On Wednesday, July 06, 2005 15:44:59 -0700 Michael Hannon <jmh AT 
physics.ucdavis DOT edu> wrote:

> Greetings.  We've had some problems with the amanda backups on our one 
> Solaris system.  We're still tweaking, so I won't yet go into details, but I 
> did want to ask a question.
> 
> We're trying to backup some relatively large file systems to an SDLT 320 tape 
> drive.  Currently the largest file system has about 60 GB of data, but the 
> trend is upward, of course.
> 
> My question relates to the detailed handling of the holding-disk space.   
> We're trying now to use a holding disk of 3 GB, with a chunksize of 1 GB.  
> Does this make sense?  I.e., these are mostly local file systems, so amanda 
> is going to be shoveling
> data pretty quickly into the holding-disk area.  There are 60 of our chunks 
> in the large file system, while our holding disk can hold only 3 of our 
> chunks.  Is it possible that amanda would put chunks into the holding disk 
> faster than they can be
> removed to the tape drive (and, hence, conclude that there was no more 
> holding-disk space)?

Normally the various dumpers would be writing to holdingdisk in parallel
at the same time the taper process picks a completed dump image off the
holdingdisk and writes it to tape.  

If a backup of a DLE won't fit on your holdingdisk, Amanda will bypass
the holdingdisk and dump directly to tape.  If you have many dumps
larger than your holdingdisk, you are effectively serializing your
backups, increasing your minimum backup window.  Also, if the dumper
process can't feed your tape fast enough, your tape may have to
repeatedly stop and reposition itself, further slowing you down (and
wearing out your tapes and drive).

An extra disk (or more, since Amanda can scatter the chunks over
multiple holdingdisks) would vastly increase the performance.

Frank


> Thanks.
> 
>                                       - Mike
> -- 
> Michael Hannon            mailto:hannon AT physics.ucdavis DOT edu
> Dept. of Physics          530.752.4966
> University of California  530.752.4717 FAX
> Davis, CA 95616-8677

-- 
Frank Smith                                      fsmith AT hoovers DOT com
Sr. Systems Administrator                       Voice: 512-374-4673
Hoover's Online                                   Fax: 512-374-4501


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