freebsd stable - Aug 2006 - gvinum / FreeBSD 6.1 / stale subdisks

Greetings all --

I'm running FreeBSD 6.1-RELEASE on i386 with a stock kernel, and am 
trying to build a 4 disk RAID5 array using vinum.  The issue is that, 
once the system is rebooted after initially creating the array, the 
subdisks come up as stale.

I first started by creating a 4 subdisk array on the 4 devices, with 
each subdisk sized at 200GB.  This configuration I used to test 
performance and verify the disk subsystem.  I used UFS2 with 64K 
blocks and 8K fragments.

I then deleted the volume, plex, and subdisks in order to do some 
performance testing on one of the drives using an fdisk-style 
partition with a UFS2 filesystem.  This filesystem was created using 
the same block/fragment settings.  I ran bonnie++ on this filesystem 
and did some bulk copies to check on performance.

I then returned to gvinum to recreate my RAID5 array.  The 
configuration file I used was similar to, but not exactly the same as 
the original configuration file as I wanted to try a different stripe 
size.  The filesystem was created with the same configuration as listed above.

After running gvinum create gvinum.config, formatting the filesystem, 
and mounting it, I am able to successfully use the filesystem to read 
and write data.  However, after rebooting, the subdisks all show as 
being stale, as shown below:

gvinum -> list
4 drives:
D drive03               State: up       /dev/ad4        A: 
38474/238475 MB (16%)
D drive02               State: up       /dev/ad6        A: 
38474/238475 MB (16%)
D drive04               State: up       /dev/ad8        A: 
38474/238475 MB (16%)
D drive01               State: up       /dev/ad10       A: 
38474/238475 MB (16%)

1 volume:
V vol1                  State: down     Plexes:       1 Size:        585 GB

1 plex:
P vol1.p0            R5 State: down     Subdisks:     4 Size:        585 GB

4 subdisks:
S vol1.p0.s0            State: stale    D: drive01      Size:        195 GB
S vol1.p0.s1            State: stale    D: drive02      Size:        195 GB
S vol1.p0.s2            State: stale    D: drive03      Size:        195 GB
S vol1.p0.s3            State: stale    D: drive04      Size:        195 GB
gvinum -> quit

There is not a /var/log/vinum_history file.  grep vinum 
/var/log/messages yielded the one line below:

Aug 16 23:36:08 archive kernel: 
g_vfs_done():gvinum/vol1[READ(offset=65536, length=8192)]error = 6

Thanks in advance for any suggestions you might have.

Steve

Hello!

On Thu, 17 Aug 2006, Steve Peterson wrote:
> I'm running FreeBSD 6.1-RELEASE on i386 with a stock kernel, and am
trying to
> build a 4 disk RAID5 array using vinum.  The issue is that, once the system
> is rebooted after initially creating the array, the subdisks come up as 
> stale.
>
> 1 volume:
> V vol1                  State: down     Plexes:       1 Size:        585 GB
>
> 1 plex:
> P vol1.p0            R5 State: down     Subdisks:     4 Size:        585 GB
>
> 4 subdisks:
> S vol1.p0.s0            State: stale    D: drive01      Size:        195 GB
> S vol1.p0.s1            State: stale    D: drive02      Size:        195 GB
> S vol1.p0.s2            State: stale    D: drive03      Size:        195 GB
> S vol1.p0.s3            State: stale    D: drive04      Size:        195 GB
> gvinum -> quit
  Try to issue 'start' command against your volume:

gvinum start vol1

It seems that gvinum requires this action for newly-created volumes (luckily 
only once). If so, this fact should be stressed in both gvinum(8) and
the Handbook. Also, the Handbook says that it covers the knowlege about
5.5 and 6.1 versions, so remnants of non-GEOM vinum(4) (which is defunct and
unavailable at least in 6.x) should be removed for the sake of clarity.
I'm not sure whether BUGS section in gvinum(8) should refer to the missing
functionality of defunct vinum(4).

Sincerely, Dmitry
-- 
Atlantis ISP, System Administrator
e-mail:  dmitry@atlantis.dp.ua
nic-hdl: LYNX-RIPE