Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 27 Sep 2006 14:42:34 -0400
From:      "Rex Roof" <rexroof@gmail.com>
To:        freebsd-geom@freebsd.org
Subject:   mirror/stripe freezing under FreeBSD 5.5
Message-ID:  <6afb69aa0609271142v7bb6fe4bkbc1dd588bfaa5607@mail.gmail.com>

next in thread | raw e-mail | index | archive | help
I have 6 SATA drives set up as 3 separate mirrors of 2 disks each, and
then have those 3 volumes striped together.  I originally set this up
on a FreeBSD 5.4 machine and then moved the SATA controllers and the
disks over to a newer FreeBSD 5.5 machine.   after putting
geom_mirror_load="YES"  and geom_stripe_load="YES" in my loader.conf
and rebooting, the drives and the volumes appeared and my data came up
as I expected.

But now, I'm getting intermittent freezing on this machine.  before
adding these disks, this machine functioned fine as my workstation
with the same OS install for a few months.    During the freezing, I'm
not seeing any dmesg errors, but once I changed
kern.geom.mirror.debug: 1 and kern.geom.stripe.debug: 1 using sysctl,
I am seeing this on each side of the freezing:

GEOM_MIRROR[1]: Disk ad12 (device mirtwo) marked as dirty.
GEOM_MIRROR[1]: Disk ad16 (device mirthr) marked as dirty.
GEOM_MIRROR[1]: Disk ad4 (device mirone) marked as dirty.
GEOM_MIRROR[1]: Disk ad6 (device mirtwo) marked as dirty.
GEOM_MIRROR[1]: Disk ad14 (device mirthr) marked as dirty.
GEOM_MIRROR[1]: Disk ad8 (device mirone) marked as dirty.
GEOM_MIRROR[1]: Disk ad16 (device mirthr) marked as clean.
GEOM_MIRROR[1]: Disk ad14 (device mirthr) marked as clean.
GEOM_MIRROR[1]: Disk ad12 (device mirtwo) marked as clean.
GEOM_MIRROR[1]: Disk ad6 (device mirtwo) marked as clean.
GEOM_MIRROR[1]: Disk ad4 (device mirone) marked as clean.
GEOM_MIRROR[1]: Disk ad8 (device mirone) marked as clean.

I see this freezing every 10 or 20 minutes, and it remains frozen for
about 90 seconds.   I run screen on this machine, and my screen
sessions work fine, but trying to run dmesg or w or similar commands
results in blocking until the disks start responding again.

I tried increasing the debug level to 2, using sysctl, but that gave
me way too much info and started constant disk hangups.
This volume stores mostly music , accessed via mt-daapd and nfsd.
In the previous FreeBSD 5.4 machine, this disk setup worked fine.
Also, when I moved the disks from one machine to the other, neither
the sata controllers or the disks were connected in the same order,
but I was under the impression that that wouldn't matter.

Anyone have any ideas?  or any suggestions for further debugging?



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?6afb69aa0609271142v7bb6fe4bkbc1dd588bfaa5607>