Amanda-Users

Re: degraded mode

2003-12-12 10:14:33
Subject: Re: degraded mode
From: Joshua Baker-LePain <jlb17 AT duke DOT edu>
To: Brian Cuttler <brian AT wadsworth DOT org>
Date: Fri, 12 Dec 2003 10:11:51 -0500 (EST)
On Fri, 12 Dec 2003 at 9:57am, Brian Cuttler wrote

> If the holding disk fills during a run, will amanda switch to
> no-holding disk mode or will it fail partitions that haven't
> yet dumped ?

It should flush the stuff on disk to tape before it starts any dumps that 
won't fit in the available holding disk space (the size of which is set at 
the beginning of the run, I think).

> ie: situation I wrote about a couple of days ago. Two amanda
> servers running concurrently, one for local disk used what it

Ah, there's your problem.  The two configs don't know about each other...

> wanted of the /amanda/work and the server that handles (client!=server)
> found at some point that there was no more holding disk space.
> Additional partitions failed, rather than switching modes or delaying
> the dumps temporarilly. Amanda will not make the switch mid-run ?

Not that I'm aware of.  As I understand it, amanda figures out how much 
disk space is available to it at the beginning of the run (when reading 
amanda.conf), and then keeps track of how much it stores there from then 
on by subtracting and adding to that number.  So, if you have two configs 
running simultaneously and don't configure them to leave space for each 
other, one or the other may unexpectedly run out of disk space (i.e. it 
thought it had 20GB left, but didn't realize the other config had already 
written 20GB).  Since it didn't anticipate this, it's an error situation, 
and the dump FAILs.

I'll be doing this soon (two configs simultaneously), and I was planning 
on configuring hard limits on the amounds of holding disk each config 
could use in order to avoid exactly this situation.

In your situation, I would setup the config that backs up the server 
itself to not use the holding disk.  On a local only config, it doesn't 
really buy you much speed, so why bother?

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University


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