RE: dump question...

2000-11-28 Thread Bort, Paul
Title: RE: dump question...





I thought I saw the 'fill the tape' explanation in the docs somewhere, I will go perusing. Your explanation is consistent with the behavior I see on my backups, though.

-Original Message-
From: Jens Bech Madsen [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, November 28, 2000 10:21 AM
To: amanda user
Subject: Re: dump question...



Bort, Paul [EMAIL PROTECTED] writes:


 Each tape contains a mixture of full and incremental backups. AMANDA picks the levels of backup for each disk by trying to

 fit as much data on tape after all of the required full backups are done. This minimizes tape hunting when you need to

 restore a file and also makes your backup runs a little more predictable in duration.


It fits as much as possible to a tape? My experience is that it tries
to make the backups equally big each time so the duration is
predictable. My tapes certainly aren't filled every day.


/Jens
-- 
Jens Bech Madsen
The Stibo Group, Denmark





RE: dump question...

2000-11-28 Thread Eric Wadsworth

I want to offer a huge *Thanks* to Paul for writing that email. It
answered a lot of my questions. Particularly, the below paragraph is
useful... This is exactly what I needed to know right now! I'm about to
add amdump to cron. Thanks, Paul!  Eric

On Tue, 28 Nov 2000, Bort, Paul wrote:
 Generally, once you have a list of disks you want to back up, they should be
 added to the disklist gradually over the course of dumpcycle backups, so
 that one tape doesn't have to try to hold too many level 0 backups at first.
 (AMANDA always starts with a level 0 for a new disk.) 





Re: Dump question

2000-11-27 Thread John R. Jackson

Should a 900MB dump over a 100Mbps network take about 10-12 hours to
complete?  ...

This could be caused by several things, such as disk contention, disk
fragmentation, disk bad blocks, SCSI problems, etc.

However, whenever I see the term "100 Mbps" (I assume you mean EtherNet)
and slow performance, I instinctively start checking for duplex problems,
and 99% of the time this turns out to be the problem.  Solaris, in
particular, is notorious for not auto-negotiating it properly for their
hme interface, resulting in a speed drop of several orders of magnitude.
The magic /etc/system incantations have been posted here several times,
or if you can't find them, E-mail me offline.

Shane T. Ferguson

John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]