Messages in this thread | | | From | "Roeland M.J. Meyer" <> | Subject | RE: LOTS OF BAD STUFF in raid0: raid0145-19990824-2.2.11 is unstable | Date | Sun, 7 Nov 1999 09:16:45 -0800 |
| |
Uh oh! This is begining to smell like some sort of "race" condition.
> -----Original Message----- > From: owner-linux-raid@vger.rutgers.edu > [mailto:owner-linux-raid@vger.rutgers.edu]On Behalf Of Mike Black > Sent: Sunday, November 07, 1999 4:53 AM > To: P.A.M. van Dam; Florian Lohoff > Cc: linux-raid@vger.rutgers.edu; linux-kernel@vger.rutgers.edu > Subject: Re: LOTS OF BAD STUFF in raid0: raid0145-19990824-2.2.11 is > unstable > > > I had problems with the "beyond end of device" message quite > regularly until > I switched the device from non-RAID to RAID and replaced the CPU. > > I was able to reproduce this with an application that was > extremely heavy > I/O on a single SCSI disk (this was linear file access on about 20G of > data). This was on a dual PIII-450. Once I switched to a > 3x50G RAID device > the problems went away ( I even had one of the "beyond" > messages while I was > copying). However, I then started having lockups instead of > "beyond" -- and > I finally traced it down to one of the CPUs (turns out one of them was > overheating). When I replaced the PIII-450's with two new > PIII-500's the > system has been rock solid since (I'm now running 2.2.13 with > the 2.2.11 > RAID patch compiled on gcc 2.95-2) > > I have another box that was Non-SMP (Kernel is SMP though) > PII/233 IDE that > I saw the "beyond end of device" message once. So, I think > that eliminates > SCSI, RAID, and SMP as likely candidates. > > So, I think there really is a bug (or maybe just a behaviour) > lurking in the > i/o buffers that just gets exacerbated by other problems, > heavy I/O, cache > problems (like overheated CPU), cables, etc. > > Maybe we could put a "marker" in the buffer_head at certain > points to see if > it gets walked on? > ----- Original Message ----- > From: P.A.M. van Dam <nucleus@ramoth.xs4all.nl> > To: Florian Lohoff <flo@rfc822.org> > Cc: David Mansfield <david@cobite.com>; <linux-raid@vger.rutgers.edu>; > <linux-kernel@vger.rutgers.edu>; Ingo Molnar > <mingo@chiara.csoma.elte.hu>; > Alan Cox <alan@lxorguk.ukuu.org.uk> > Sent: Saturday, November 06, 1999 2:40 PM > Subject: Re: LOTS OF BAD STUFF in raid0: raid0145-19990824-2.2.11 is > unstable > > > On Sat, Nov 06, 1999 at 03:44:47PM +0100, Florian Lohoff wrote: > > > > Easy setup > > > > raid0 of raid5s ... > > > > read: > > > > /dev/md0 raid0 /dev/md1 /dev/md2 > > /dev/md1 raid5 scsi disks (tried with 6 and 3) > > /dev/md2 raid5 scsi disks (tried with 6 and 3) > > > > As soon as you initiate a "mke2fs /dev/md0" the machine > > hangs with "Got md request, not good" (md.c -> do_md_request()) > > Same here, but then with Mauelschagens LVM. An mkfs -t ext2 for the > first time (after bootup) on a striped LV will trigger the "Got LVM > request, not good" message (and accompanied by some "trying to ... > beyond end of device" messages). The second attempt will not generate > these messages. > > I'm sorry, but I still have the strong belief that this isn't hardware > nor RAID (only) related. There must be some bug lurking, we have to > stamp out before we can take 2.4 seriously. > > > > > > I reproduced this with different hardware (machines, > harddrives, scsi > controllers etc - Even on sparc ...) > > Ok. Does SMP be the common denominator in your test? Almost > all reports > are with SMP machines. > > > > > >From the manpage this is a feature which should work. > > Best regards, > > Pascal > > > > > Flo > > -- > > Florian Lohoff flo@rfc822.org +49-5241-470566 > > ... The failure can be random; however, when it does occur, it is > > catastrophic and is repeatable ... Cisco Field Notice > > > > > > - > > To unsubscribe from this list: send the line "unsubscribe > linux-kernel" in > > the body of a message to majordomo@vger.rutgers.edu > > Please read the FAQ at http://www.tux.org/lkml/ >
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |