Replacing a dead disk in a zpool
After digging endlessly this night I finally found the solution. The short answer is that you can use the disks' GUIDs (which persist even after disconnecting a drive) with the zpool
command.
Long answer:
I got the disk's GUID using the zdb
command which gave me the following output
root@zeus:/dev# zdb
hermes:
version: 28
name: 'hermes'
state: 0
txg: 162804
pool_guid: 14829240649900366534
hostname: 'zeus'
vdev_children: 1
vdev_tree:
type: 'root'
id: 0
guid: 14829240649900366534
children[0]:
type: 'raidz'
id: 0
guid: 5355850150368902284
nparity: 1
metaslab_array: 31
metaslab_shift: 32
ashift: 9
asize: 791588896768
is_log: 0
create_txg: 4
children[0]:
type: 'disk'
id: 0
guid: 11426107064765252810
path: '/dev/disk/by-id/ata-ST3300620A_5QF0MJFP-part2'
phys_path: '/dev/gptid/73b31683-537f-11e2-bad7-50465d4eb8b0'
whole_disk: 1
create_txg: 4
children[1]:
type: 'disk'
id: 1
guid: 15935140517898495532
path: '/dev/disk/by-id/ata-ST3300831A_5NF0552X-part2'
phys_path: '/dev/gptid/746c949a-537f-11e2-bad7-50465d4eb8b0'
whole_disk: 1
create_txg: 4
children[2]:
type: 'disk'
id: 2
guid: 7183706725091321492
path: '/dev/disk/by-id/ata-ST3200822A_5LJ1CHMS-part2'
phys_path: '/dev/gptid/7541115a-537f-11e2-bad7-50465d4eb8b0'
whole_disk: 1
create_txg: 4
children[3]:
type: 'disk'
id: 3
guid: 17196042497722925662
path: '/dev/disk/by-id/ata-ST3200822A_3LJ0189C-part2'
phys_path: '/dev/gptid/760a94ee-537f-11e2-bad7-50465d4eb8b0'
whole_disk: 1
create_txg: 4
features_for_read:
The GUID I was looking for is 15935140517898495532
which enabled me to do
root@zeus:/dev# zpool offline hermes 15935140517898495532
root@zeus:/dev# zpool status
pool: hermes
state: DEGRADED
status: One or more devices has been taken offline by the administrator.
Sufficient replicas exist for the pool to continue functioning in a
degraded state.
action: Online the device using 'zpool online' or replace the device with
'zpool replace'.
scan: scrub repaired 0 in 2h4m with 0 errors on Sun Jun 9 00:28:24 2013
config:
NAME STATE READ WRITE CKSUM
hermes DEGRADED 0 0 0
raidz1-0 DEGRADED 0 0 0
ata-ST3300620A_5QF0MJFP ONLINE 0 0 0
ata-ST3300831A_5NF0552X OFFLINE 0 0 0
ata-ST3200822A_5LJ1CHMS ONLINE 0 0 0
ata-ST3200822A_3LJ0189C ONLINE 0 0 0
errors: No known data errors
and then
root@zeus:/dev# zpool replace hermes 15935140517898495532 /dev/disk/by-id/ata-ST3500320AS_9QM03ATQ
root@zeus:/dev# zpool status
pool: hermes
state: DEGRADED
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scan: resilver in progress since Sun Jun 9 01:44:36 2013
408M scanned out of 419G at 20,4M/s, 5h50m to go
101M resilvered, 0,10% done
config:
NAME STATE READ WRITE CKSUM
hermes DEGRADED 0 0 0
raidz1-0 DEGRADED 0 0 0
ata-ST3300620A_5QF0MJFP ONLINE 0 0 0
replacing-1 OFFLINE 0 0 0
ata-ST3300831A_5NF0552X OFFLINE 0 0 0
ata-ST3500320AS_9QM03ATQ ONLINE 0 0 0 (resilvering)
ata-ST3200822A_5LJ1CHMS ONLINE 0 0 0
ata-ST3200822A_3LJ0189C ONLINE 0 0 0
errors: No known data errors
After resilvering had been completed everything worked well again. It would have been nice to include this information, that you can use a disk's GUID obtained through zdb
with the zpool
command, with the manpage of zpool.
Edit
As pointed out by durval below the zdb
command may not output anything. Then you may try to use
zdb -l /dev/<name-of-device>
to explicitly list information about the device (even if it already is missing from the system).
The issue is the disks are referenced by ids and not by device.
Here is a workaround that should work:
ln -s /dev/null /dev/ata-ST3300831A_5NF0552X
zpool export hermes
zpool import hermes
zpool status
# note the new device name that should appear here
zpool offline hermes xxxx
zpool replace hermes xxxx /dev/disk/by-id/ata-ST3500320AS_9QM03ATQ
Edit: I was 30 seconds late ...
@Marcus: Thanks for posting this excellent answer to your own question, it helped me a lot.
The other day I found a twist that might interest you (and anyone else that comes here a-googling in the future): I had a cache device that was dropped from the pool (and marked as "UNAVAIL") due to this same error (ZFS-8000-4J, "label is missing or invalid"), and trying to offline/remove/replace it failed with exactly the same "no such device in pool" message.
BUT, when I tried to apply your solution, plain "zdb" (with no arguments) did not list the device, much less its GUID.
After some digging, I found that "zdb -l /dev/DEVICENAME" listed the GUID (taking it directly from the device, and not from the pool records), and using that GUID enabled me to do the replacement (actually I did a "zpool offline" followed by a "zpool remove" and then a "zpool add", which worked perfectly).