Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 27 Jun 1999 09:14:53 +0200
From:      Bernd Walter <ticso@cicely.de>
To:        Greg Lehey <grog@lemis.com>
Cc:        Ollivier Robert <roberto@keltia.freenix.fr>, freebsd-hackers@FreeBSD.ORG
Subject:   Re: Volume managers (was: ufs/ffs resize?)
Message-ID:  <19990627091453.A38541@cicely8.cicely.de>
In-Reply-To: <19990627093345.O427@freebie.lemis.com>; from Greg Lehey on Sun, Jun 27, 1999 at 09:33:45AM %2B0930
References:  <199906252100.OAA31969@sigma.veritas.com> <19990626195554.A37269@cicely8.cicely.de> <19990627003554.A25754@keltia.freenix.fr> <19990627093345.O427@freebie.lemis.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Jun 27, 1999 at 09:33:45AM +0930, Greg Lehey wrote:
> On Sunday, 27 June 1999 at  0:35:54 +0200, Ollivier Robert wrote:
> > I think one of the difficulty of growing a FS is that you have to
> > choose whether you need the FS to be contiguous or not. The latter
> > case makes it much more difficult...
> 
> Why shouldn't it be contiguous?  That's what the volume manager's
> there for.
> 
It should be always possible to add some blocks to an existing volume
without a gap between.
The problem is shrinking.
You usually want to shrink a fs not because you want it smaller but you want
to free a harddisk.
Murphy says that this disk is at the beginning or the middle of the partition.
If Murphy had a good day he says this disk is part of a stripe...

What system should handle this case?
For FFS it might be better handled by the volume-managaer because it should
be difficult to implement GAP-Handling in FFS
LFSs daily job is to do such jobs as a kind of garbadge collection so it should
be much faster and easier if handled by the fs - but that would mean the fs
need to know about the volumes details.

There are some other points left with interaction of volumemanagers and
Filesystem instead of shrinking and growing.
Say you have a concatenated volume with 3 disks.
One drive fails without any chance to get it up again.
Vinum takes the plex and because it's the only one the volume too.
There no kind of emergency mode in which you can say - bring that volume
up with a large defektive area in them so I can mount read-only and try
to read everthing I can.
Instead of CCD vinum knows how big the failed disk was.

One other point is that it is not possible with vinum or ccd to add space to
a striped/raid5 volume.
One way might be to use ccd to concat striped vinum volumes - never tried.
But in my opinion that's not the way it should be.

-- 
B.Walter                  COSMO-Project              http://www.cosmo-project.de
ticso@cicely.de                                      info@cosmo-project.de



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19990627091453.A38541>