xfs
[Top] [All Lists]

Re: Linux + XFS + SCSI = Problems?

To: "Gonyou, Austin" <austin@xxxxxxxxxxxxxxx>, linux-xfs@xxxxxxxxxxx
Subject: Re: Linux + XFS + SCSI = Problems?
From: cradeke <chrad@xxxxxxxxxxxxxxxxxx>
Date: Thu, 01 Nov 2001 21:25:19 +0100
References: <85063BBE668FD411944400D0B744267A8886CA@AUSMAIL>
Sender: owner-linux-xfs@xxxxxxxxxxx
"Gonyou, Austin" wrote:

> Can SOMEONE please test this and help out? I'm using pretty generic
> hardware/software config here. This is a very critical issue. Please help if
> you can just take 15mins to run the scripts.
>

sorry, but I can't do it because my data on disk are important for me.. I also
use a dual machine (p3@asus p2b-d) and had the same problem sometimes... I
heard that there is a problem with kupdatd... A.C. say somewhere in a kernel
mailinglist "it is mysteriuos.. no solution at the moment".. the problem exists
since the kernel is SMP-able... what it caused is not easy to say.. for me it
occures since I work with ramfs enabled in the kernel and if I use a scsi
scannner too... I could not find out when exactly the problem happens. but at
the moment the system seems save: no scanner is connected and the ramfs is not
touched/mounted and all goes well for about 3 weeks uptime

[cradeke@usher radeke]$ uname -a
Linux usher 2.4.5-xfs-1.0.1 #7 SMP Wed Oct 31 20:50:40 CET 2001 i686 unknown

is the actual kernel.. I saw it with xfs-1.0 and 1.0.1 too

may that helps a bit

regards

c.radeke


>
> --
> Austin Gonyou
> Systems Architect, CCNA
> Coremetrics, Inc.
> Phone: 512-796-9023
> email: austin@xxxxxxxxxxxxxxx
>
> > -----Original Message-----
> > From: Gonyou, Austin [mailto:austin@xxxxxxxxxxxxxxx]
> > Sent: Wednesday, October 31, 2001 4:29 PM
> > To: Linux XFS Mailing List
> > Subject: RE: Linux + XFS + SCSI = Problems?
> >
> >
> > I just started trying it. Sorry for the conusion. I started
> > using it because
> > I put a sleep(1); just after the print garbage and noticed
> > that the system
> > was writing out in 4096 byte chunks. The mount options there
> > are from a
> > progression of trying different things.
> >
> > --
> > Austin Gonyou
> > Systems Architect, CCNA
> > Coremetrics, Inc.
> > Phone: 512-796-9023
> > email: austin@xxxxxxxxxxxxxxx
> >
> > > -----Original Message-----
> > > From: Steve Lord [mailto:lord@xxxxxxx]
> > > Sent: Wednesday, October 31, 2001 3:45 PM
> > > To: Gonyou, Austin
> > > Cc: Linux XFS Mailing List
> > > Subject: Re: Linux + XFS + SCSI = Problems?
> > >
> > >
> > > On Wed, 2001-10-31 at 15:37, Gonyou, Austin wrote:
> > > > I'm trying to uncover the root of an issue.
> > > > The issue is this:
> > > >   When running a perl script I can get perl to jump up to
> > > 99% cpu. When this
> > > > happens, I can't kill the process.
> > > > The only thing I can do to recover properly is reset the
> > > system. Either
> > > > power cycle or reset.
> > > > Is there anyone who could attempt to run the attatched
> > > script and let me
> > > > know if they experience something similar?
> > > >
> > > > The targe configuration of the box I'm most concerned with
> > > is as follows:
> > > > 512MB RAM (ecc, parity, registered)
> > > > 1x 9GB SCSI HDD(seagate)
> > > > 1x PCI RAID controller(AMI MegaRAID express) wrthru and
> > read-cached.
> > > > DUAL 550 PIII
> > > > RH 7.1 SGI XFS 1.0 installed or later.
> > > > non-updated perl 5.6.0(installed with 7.1)
> > > > Mount options for the filesystem I'm testing with.
> > > > /dev/sda5 on /home type xfs (rw,biosize=13,logbufs=8,osyncisdsync)
> > > >
> > > > 2.4.13 Vanilla kernel + SGI XFS patch for 2.4.13.
> > > > Attatched are the perl scripts I use to test with. Also the
> > > kernel config
> > > > I'm using. I'm pretty desparate here.
> > > > If you can offer help on this, please help!
> > > >
> > > >
> > > > --
> > > > Austin Gonyou
> > > > Systems Architect, CCNA
> > > > Coremetrics, Inc.
> > > > Phone: 512-796-9023
> > > > email: austin@xxxxxxxxxxxxxxx
> > > >
> > > > ----
> > > >
> > >
> > > Austin, you did not tell me you were using biosize, have you tried
> > > without it?
> > >
> > > Steve
> > >
> > > --
> > >
> > > Steve Lord                                      voice:
> > +1-651-683-3511
> > > Principal Engineer, Filesystem Software         email: lord@xxxxxxx
> > >
> >


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