Date: Mon, 3 Sep 2001 22:56:19 +0200 (CEST)
From: Seth Mos <knuffie@xs4all.nl>
To: "Gonyou, Austin" <austin@coremetrics.com>
cc: "'XFS mailing list'" <linux-xfs@oss.sgi.com>
Subject: RE: System lock while accessing files causes file corruption
In-Reply-To: <85063BBE668FD411944400D0B744267A888527@AUSMAIL>
Message-ID: <Pine.BSI.4.10.10109032253100.15247-100000@xs3.xs4all.nl>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-linux-xfs@oss.sgi.com
Precedence: bulk

On Mon, 3 Sep 2001, Gonyou, Austin wrote:

> One quick not on this, is there anything I can do kernel wise to prevent
> this without striking a crapload of overhead on the system?

No, all metadata journaling filesystems have this. ReiserFS has it and I
suspect that JFS can do this as well. If I ever get the NCR MP-RAS box to
be out of production I could test this with Veritas as well.

I hate to say it but this a general metadata jfs problem/feature. It needs
a smarter kernel to make it go to disk faster when the box is not loaded.

> > -----Original Message-----
> > From: Gonyou, Austin 
> > Sent: Monday, September 03, 2001 12:41 PM
> > To: 'Seth Mos'; Gonyou, Austin; XFS mailing list
> > Subject: RE: System lock while accessing files causes file corruption
> > 
> > 
> > That's all well and good..but what about the configuration 
> > files? They are text and are not redundant in the same way. I 
> > know it's been done to death..and I've read a lot of this 
> > stuff, but even after making thanges, etc, especially if I 
> > use logbufs > 2, I can make this happen at will. That's why 
> > I've got concerns. I know about the points Keith made, and 
> > very valid in this case especially, but my major concern is 
> > deploying a TB size db only to get taken down that my configs 
> > are messed up. (yes, I know that's what backups and CVS are 
> > for) but that's not acceptable when talking about my primary 
> > FS. I'm going to go through as many iterations of this as I 
> > can to see if I can narrow down exactly what and where, what 
> > hardware, etc. We're about to do a major deployment and I'm 
> > just trying to do due dilligence, beyond the FAQ and random 
> > mails. Thanks for listening, and this can certainly come 
> > off-list if you feel that's best. I really don't want to beat 
> > a dead horse anymore than I have to. Assurance of success in 
> > this respect is very important. We don't have a lot of 
> > storage to throw around. 
> > 
> > -- 
> > Austin Gonyou
> > Systems Architect, CCNA
> > Coremetrics, Inc.
> > Phone: 512-796-9023
> > email: austin@coremetrics.com 
> > 
> > > -----Original Message-----
> > > From: Seth Mos [mailto:knuffie@xs4all.nl]
> > > Sent: Sunday, September 02, 2001 7:18 PM
> > > To: Gonyou, Austin; XFS mailing list
> > > Subject: Re: System lock while accessing files causes file 
> > corruption
> > > 
> > > 
> > > At 18:16 2-9-2001 -0500, Gonyou, Austin wrote:
> > > >Why is this? If I open a file, text/otherwise and the power 
> > > actually fails,
> > > >(i turn it off), once in a while I get a corrupt file. Why 
> > > is this? What
> > > >would happen if I was writing to some Oracle filesystems and 
> > > this situation
> > > >occurred? Please advise.
> > > 
> > > See the http://oss.sgi.com/projects/xfs/faq.html#nulls
> > > 
> > > A database would survive since most have their own buffering and 
> > > transaction scheme.
> > > 
> > > 
> > > --
> > > Seth
> > > Every program has two purposes one for which
> > > it was written and another for which it wasn't
> > > I use the last kind.
> > > 
> > 
> 
>