Amanda-Users

Re: amflush problem

2004-10-15 16:02:08
Subject: Re: amflush problem
From: Frederic Medery <dist-list AT LEXUM.UMontreal DOT CA>
Date: Fri, 15 Oct 2004 14:06:39 -0400
[root@salerno daily]# du -sh *
11G     20040823
117M    20040910
31G     20041012
25G     20041014


Our drive is using LTO ultrium 100/200  GB cartridge + tape

Frederic Medery
System Administrator

LexUM, University of Montreal



Gene Heskett wrote:

I see that you are using a 1GB chunk size setting, and just a rather cursory mental addup of one of them indicates it would take about a 21 to 22GB tape to actually hold just the dump of

pisa.lan.lexum.pri._home_projets

Are your tapes that big? We apparently neglected to specifically ask that.

Since amanda cannot span an individual dump (all those chunks added together again) over more than one tape, it probably totals things up and throws up its hands in defeat since the backup is bigger than the tape. The design of amanda tends to want its disklist to be rather lengthy, and broken up such that no one backup image is more than say 10% of a tape, and with a few that may be only a few megabytes, which can then allow amanda to juggle its scheduleing to facilitate a consistent amount of data (on the tape, after any gzip compression if used) each night. I have had it fill the tapes to 97+% many nights in a row without ever hitting the dreaded EOT here.

Unless the budget has the sheckles for a bigger drive, I think I'd be rather inclined to rm -fR all of this, and breakup the disklist entries into smaller sized pieces and more of them. Start with the biggest one tonight, making 4 or 5 pieces (disklist entries) out of it, do the next biggest one tomorrow night, etc till you have it carved down into gigabyte or smaller hunks per entry. Or get a larger drive, much larger. See sheckles above...


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