Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 18 Nov 2004 23:49:21 -0500
From:      Paul Mather <paul@gromit.dlib.vt.edu>
To:        freebsd-geom@freebsd.org
Subject:   geom_mirror synchronisation hangs at boot
Message-ID:  <1100839762.5421.21.camel@zappa.Chelsea-Ct.Org>

next in thread | raw e-mail | index | archive | help
I had another one of those occasional "TIMEOUT - WRITE_DMA" messages.
This time it was on a 6.0-CURRENT system (last built 2004-11-12) with a
geom_mirror setup, and it caused one of the providers to be removed from
the mirror and to operate in degraded mode:

Nov 18 14:15:07 zappa kernel: ad0: TIMEOUT - WRITE_DMA retrying (2 retries left) LBA=49981679
Nov 18 14:15:10 zappa kernel: ad0: FAILURE - WRITE_DMA timed out
Nov 18 14:15:10 zappa kernel: GEOM_MIRROR: Cannot update metadata on disk ad0 (error=5).
Nov 18 14:15:10 zappa kernel: GEOM_MIRROR: Device raid1: provider ad0 disconnected.
Nov 18 14:15:10 zappa kernel: GEOM_MIRROR: Request failed (error=1). ad0[WRITE(offset=809639936, length=2048)]


As with my geom_vinum episode, the drive wasn't really down.  So, like
my geom_vinum case, I decided to reboot to make the drive magically be
recognised as alive.  (In retrospect, an atacontrol detach/attach would
have been better.  I'll remember that in future.:)

When the system rebooted, all appeared well.  The drive was recognised
and I could hear reconstruction kick in as ad0 was rebuilt.
Unfortunately, a moment later, when it came to probe my atapicam devices
(cd0 and da0 ZIP drive), the sounds of furious reconstruction vanished
and the machine locked up.

I am wondering if the atapicam device probing interfered with/was
affected by the rebuilding that had kicked off on ad0.  Is there some
way of delaying reconstruction such that it begins after some timed
delay---rather like the way the background fsck is delayed for 60
seconds to get most of the boot sequence completed before it begins
pounding on the drives?  I've never yet had any problems reconstructing
a mirror when the OS is up and running.

In the end, the only way I seemed to be able to render my system
bootable was to download and burn the latest FreeSBIE BETA CD, boot
that, and then "kldload geom_mirror" to initiate reconstruction once the
system was running. :-(

I know there are various gmirror options to disable
auto-synchronisation.  Is there anything that can be twiddled in the
loader to enable/disable auto-synchronisation?  That would be really
handy.

Here is the typical order of probing of ATA devices in my system.  Note
that cd0 and da0 come after the GEOM_MIRROR discovery/initialisation:

Nov 18 19:36:44 zappa kernel: ad0: 24405MB <IBM-DJNA-352500/J51OA30K> [49585/16/63] at ata0-master UDMA33
Nov 18 19:36:44 zappa kernel: acd0: DVDR <LITE-ON DVDRW SOHW-832S/VS08> at ata0-slave UDMA33
Nov 18 19:36:44 zappa kernel: ata1-slave: FAILURE - SETFEATURES SET TRANSFER MODE status=1<ERROR> error=4<ABORTED>
Nov 18 19:36:44 zappa kernel: ad2: 24405MB <IBM-DJNA-352500/J51OA30K> [49585/16/63] at ata1-master UDMA33
Nov 18 19:36:44 zappa kernel: afd0: REMOVABLE <IOMEGA ZIP 100 ATAPI/23.D> at ata1-slave BIOSPIO
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1 created (id=1030107361).
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1: provider ad0 detected.
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1: provider ad2 detected.
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1: provider ad2 activated.
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1: provider ad0 activated.
Nov 18 19:36:44 zappa kernel: GEOM_MIRROR: Device raid1: provider mirror/raid1 launched.
Nov 18 19:36:44 zappa kernel: da0 at ata1 bus 0 target 1 lun 0
Nov 18 19:36:44 zappa kernel: da0: <IOMEGA ZIP 100 23.D> Removable Direct Access SCSI-0 device 
Nov 18 19:36:44 zappa kernel: da0: 3.300MB/s transfers
Nov 18 19:36:44 zappa kernel: da0: 96MB (196608 512 byte sectors: 64H 32S/T 96C)
Nov 18 19:36:44 zappa kernel: cd0 at ata0 bus 0 target 1 lun 0
Nov 18 19:36:44 zappa kernel: cd0: <LITE-ON DVDRW SOHW-832S VS08> Removable CD-ROM SCSI-0 device 
Nov 18 19:36:44 zappa kernel: cd0: 33.000MB/s transfers
Nov 18 19:36:44 zappa kernel: cd0: Attempt to query device size failed: NOT READY, Medium not present
Nov 18 19:36:44 zappa kernel: Mounting root from ufs:/dev/mirror/raid1a

In the boot hang on auto-synchronisation, the hang would occur just
before the "Mounting root" line.

Cheers,

Paul.
-- 
e-mail: paul@gromit.dlib.vt.edu

"Without music to decorate it, time is just a bunch of boring production
 deadlines or dates by which bills must be paid."
        --- Frank Vincent Zappa



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1100839762.5421.21.camel>