Odd behavior with volume labels and gmirror
Ben Schumacher
me at benschumacher.com
Fri Jan 8 18:07:16 UTC 2010
Hey-
I haven't search the list, so this could be covered already, but I've
seen odd behavior with volume labels and gmirror.
First some background.
Here's my uname:
$ uname -a
FreeBSD filer.sats.internal 8.0-RELEASE-p1 FreeBSD 8.0-RELEASE-p1 #1:
Mon Dec 21 10:21:37 MST 2009
ben at filer.sats.internal:/usr/obj/data/FreeBSD/RELENG_8_0/src/sys/FILER
i386
I have a 4-way gmirror:
$ gmirror status
Name Status Components
mirror/root COMPLETE ad6s1
ad12s1
ad14s1
ad18s1
With multiple labeled slices:
$ tunefs -p /dev/mirror/roota
tunefs: ACLs: (-a) disabled
tunefs: MAC multilabel: (-l) disabled
tunefs: soft updates: (-n) disabled
tunefs: gjournal: (-J) disabled
tunefs: maximum blocks per file in a cylinder group: (-e) 2048
tunefs: average file size: (-f) 16384
tunefs: average number of files in a directory: (-s) 64
tunefs: minimum percentage of free space: (-m) 8%
tunefs: optimization preference: (-o) time
tunefs: volume label: (-L) root
$ tunefs -p /dev/mirror/rootf
tunefs: ACLs: (-a) disabled
tunefs: MAC multilabel: (-l) disabled
tunefs: soft updates: (-n) enabled
tunefs: gjournal: (-J) disabled
tunefs: maximum blocks per file in a cylinder group: (-e) 2048
tunefs: average file size: (-f) 16384
tunefs: average number of files in a directory: (-s) 64
tunefs: minimum percentage of free space: (-m) 8%
tunefs: optimization preference: (-o) time
tunefs: volume label: (-L) usr
And used these in my /etc/fstab for mounting:
$ cat /etc/fstab
# Device Mountpoint FStype Options Dump Pass#
/dev/mirror/rootb.eli none swap sw 0 0
/dev/da0b none swap sw 1 0
/dev/ufs/root / ufs rw 1 1
/dev/ufs/tmp /tmp ufs rw,nosuid 2 2
/dev/ufs/usr /usr ufs rw 2 2
/dev/ufr/var /var ufs rw 2 2
The issue I experienced last night is that after a power outage, the
system detected the labelled filesystems names from some of the GEOM
providers of the mirror -- specifically they appear to all be mapped
to "/dev/ad18s1[adef]" and the machine never managed to fully boot
without some manual intervention to change /etc/fstab back to using
the mirror devices instead of the labels.
Insult to injury, I *may* have a drive going bad (ad18), but given
what was happening with the mounting, I'm not really sure. I used
recoverdisk to dup that disk to the others and it seemed to work fine,
but during the label problem I was getting some READ_DMA errors trying
to rebuild the mirror. I guess my question is, is using labels on
mirrors unreliable since the labels write through and apparently can
get detected from the devices before the mirror is loaded? Or is this
a bug that I should file a pr for?
Thanks for your attention.
Cheers,
Ben
More information about the freebsd-fs
mailing list