Because XFS is BETTER (Re: opinion on XFS)

Donald Becker becker at
Wed May 8 18:48:29 PDT 2002

On Thu, 9 May 2002, Eray Ozkural wrote:

> On Thursday 09 May 2002 02:15, Mark Hahn wrote:
> >
> > in short, what is it that XFS offers that people want so badly?
> Your nodes don't die. That simple.
> We are not so stupid to use crippled hardware. We're using exactly the same
> hardware before switching to XFS.

With SCSI disks or IDE disks?

In the months following the initial release of the Linux-XFS kernel
(based on 2.3.99), we followed it closely.  Every time we would load a
new kernel on a fresh system, XFS would horribly corrupt itself.  This
continued for many months.  While at OLS I asked a XFS developer why
they released a kernel that could not even boot without corrupting
itself.  It turns out that they had never tested with IDE disks, and
were using the IDE driver interface in an invalid way.  (When you change
the block driver interface, it's difficult to claim that the bug is in
the IDE driver.)  It supposedly worked just great with SCSI disks.

> How come you haven't heard many horror stories with EXT2? Almost
> anybody who's done some heavy coding or busy servers on linux would
> know that EXT2 is very likely to cause you more headaches, downtime,
> and corruption, than, say, FAT filesystem.

I've developed Linux drivers for almost a decade, and never lost a ext2
filesystem through many horrible kernel crashes.  There were problems
with many of the 2.3.99* and a few of the early 2.4 kernels, but those
were not caused by the ext2 code.

Donald Becker				becker at
Scyld Computing Corporation
410 Severn Ave. Suite 210		Second Generation Beowulf Clusters
Annapolis MD 21403			410-990-9993

More information about the Beowulf mailing list