[Top] [All Lists]

Re: XFS for postgres databases?

To: Nathan Scott <nathans@xxxxxxx>
Subject: Re: XFS for postgres databases?
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Wed, 12 May 2004 13:00:47 +0100
Cc: Nigel Cunningham <ncunningham@xxxxxxxxxxxxx>, stevew@xxxxxxxxxxxxxxx, "Jeffrey W. Baker" <jwbaker@xxxxxxx>, linux-xfs@xxxxxxxxxxx
In-reply-to: <20040512165624.C389759@xxxxxxxxxxxxxxxxxxxxxxxx>; from nathans@xxxxxxx on Wed, May 12, 2004 at 04:56:24PM +1000
References: <200405121300.14866.stevew@xxxxxxxxxxxxxxx> <1084332608.11308.3.camel@noodles> <20040512143941.A389759@xxxxxxxxxxxxxxxxxxxxxxxx> <200405121618.27843.ncunningham@xxxxxxxxxxxxx> <20040512165624.C389759@xxxxxxxxxxxxxxxxxxxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mutt/
On Wed, May 12, 2004 at 04:56:24PM +1000, Nathan Scott wrote:
> Christoph has been working on fixing that up in 2.6, he'll have the
> gory details if you're curious.  (actually, Christoph/Steve - do we
> have to use bdev->bd_inode->i_mapping as the buftarg mapping?  what
> if we fudged that mapping, then our pages wouldn't overlap with the
> block device inode pages - wouldn't that work around this?  maybe
> Russells hack^Widea of using multiple metadata mappings to overcome
> the pagecache size limits would be helped there too...?)

We can use a differnt mapping, in fact that's what XFS did before 2.4.10.
It's a little bit of additional code, but not a big deal.  The real problem
with that is that now the blockdev mapping and XFS mapping are compltely
unsynchonized, e.g. reading from the blockdev will give you stale data.
For xfs_db this means it'll get access to an totally incoherent image of
the filesystem.

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