Hi, all,
during the last couple of years I occasionally increased the capacity
of raidz2 based zpools by replacing one disk at a time and resilvering,
subsequently. After replacing the final disk and a reboot (I guess
zpool export & zpool import would have done the trick, too) the capacity
of the FS on top of that pool was increased according to the size of
the new disk. All of theses systems had a pool built on one single vdev.
Last week I exchanged all disks of one vdev that is part of a 2 vdev
zpool. According to the Solaris documentation I found that should be
possible. I always assumed vdevs were sort of independent of each
other.
My observations:
During resilvering the activity LEDs of all 12 disks were showing heavy
load, not only the ones of the 6 disks being part of the vdev in question.
After exchanging all 6 disks the capacity stayed the same. I tried
zpool export, zpool import
reboot
zpool scrub
to no avail.
datatomb2# zpool status sx40
pool: sx40
state: ONLINE
scan: scrub repaired 0 in 3h32m with 0 errors on Sat Jun 2 00:41:38 2012
config:
NAME STATE READ WRITE CKSUM
sx40 ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
gpt/sx40-disk0 ONLINE 0 0 0
gpt/sx40-disk1 ONLINE 0 0 0
gpt/sx40-disk2 ONLINE 0 0 0
gpt/sx40-disk3 ONLINE 0 0 0
gpt/sx40-disk4 ONLINE 0 0 0
gpt/sx40-disk5 ONLINE 0 0 0
raidz2-1 ONLINE 0 0 0
gpt/sx40-disk6 ONLINE 0 0 0
gpt/sx40-disk7 ONLINE 0 0 0
gpt/sx40-disk8 ONLINE 0 0 0
gpt/sx40-disk9 ONLINE 0 0 0
gpt/sx40-disk10 ONLINE 0 0 0
gpt/sx40-disk11 ONLINE 0 0 0
errors: No known data errors
datatomb2# zpool get all sx40
NAME PROPERTY VALUE SOURCE
sx40 size 10.9T -
sx40 capacity 78% -
sx40 altroot - default
sx40 health ONLINE -
sx40 guid 1478259715706579670 default
sx40 version 28 default
sx40 bootfs - default
sx40 delegation on default
sx40 autoreplace off default
sx40 cachefile - default
sx40 failmode wait default
sx40 listsnapshots off default
sx40 autoexpand on local
sx40 dedupditto 0 default
sx40 dedupratio 1.00x -
sx40 free 2.31T -
sx40 allocated 8.57T -
sx40 readonly off -
The first 6 disks building raidz2-0 are 2 TB ones, not 1 TB.
The gpt partitions *are* about 2 TB in size.
What am I missing? Any hints welcome. I do have the hardware
to build another device with 6 drives 2 TB and 6 drives 1 TB,
which I planned to hook up to another server. Of course I could
connect it to this one first, build a second pool, copy over the
data ... but I was trying to avoid that in the first place ;-)
Thanks in advance,
Patrick
--
punkt.de GmbH * Kaiserallee 13a * 76133 Karlsruhe
Tel. 0721 9109 0 * Fax 0721 9109 100
info@punkt.de http://www.punkt.de
Gf: J?rgen Egeling AG Mannheim 108285