Disk Failure in 8.3

[tdyb0815]

I just upgraded from older version to 8.3 a few days ago and everything has been working great. I had a zpool with 5 2 TB disks (4 active and 1 spare). It appears this morning that one of the drives has failed. I can see my Volume listed but cannot get a status. Also, I can view 4 disks (3 of which are member of media stripe and 1 is member of media spare), but i cannot access data. Can anyone advise on how to promote the spare to active and if that is the correct course of action?

I'm a bit of a noob so looking for some advice so i don't lose my data.

Thanks.

 

[tdyb0815]

I powered off, and put the disk back in and now it works again. But I'm sure the drive is on the verge of going bad (it has happened this way before). Trying to offline the bad disk but i keep getting the "cannot offline no valid replicas" message. Even after scrubbing the volume. Any ideas on how to offline the disk?

 

[cyberjock]

If you are getting the error that you cannot offline because there are no valid replicas you likely don't have a zpool capable of disk failure.

Can you put the output of 'zpool status' and put it in code tags please.

 

[tdyb0815]

If you are getting the error that you cannot offline because there are no valid replicas you likely don't have a zpool capable of disk failure.

Can you put the output of 'zpool status' and put it in code tags please.

pool: Media
state: ONLINE
scan: scrub repaired 0 in 0h27m with 0 errors on Tue Jan 15 08:25:22 2013
config:

NAME STATE READ WRITE CKSUM
Media ONLINE 0 0 0
ada0p2 ONLINE 0 0 0
ada1p2 ONLINE 0 0 0
ada2p2 ONLINE 0 0 0
ada3p2 ONLINE 0 0 0
spares
ada4p2 AVAIL

errors: No known data errors

 

[cyberjock]

Yeah, it looks like you have no redundancy from a disk failure. That spare disk is useless for you. If a disk fails the spare won't save you.. you'll just lose all of your data.

Here's what you need to do.. sooner than later.

1. Backup your zpool. You already have a failing disk and if it dies for good your data is also gone for good.
2. Destroy and recreate the zpool using anything with redundancy. My recommendation for 5 disks would be a RAIDZ1.
3. Copy all of the data back to the zpool.

If you have shares and whatnot already created you can destroy the zpool, recreate the zpool with the same name and folder structure. Then you'll only need to resetup your permissions and you'll be all fixed up and protected from a single disk failure.

 

[tdyb0815]

Yeah, it looks like you have no redundancy from a disk failure. That spare disk is useless for you. If a disk fails the spare won't save you.. you'll just lose all of your data.

Here's what you need to do.. sooner than later.

1. Backup your zpool. You already have a failing disk and if it dies for good your data is also gone for good.
2. Destroy and recreate the zpool using anything with redundancy. My recommendation for 5 disks would be a RAIDZ1.
3. Copy all of the data back to the zpool.

If you have shares and whatnot already created you can destroy the zpool, recreate the zpool with the same name and folder structure. Then you'll only need to resetup your permissions and you'll be all fixed up and protected from a single disk failure.

Thanks. Better to find out now than later. Would you recommend doing a ZFS Replication from within the 8.3 UI or just simply copy to another location, redo, then copy back?

 

[cyberjock]

Ehh.. uhh.. ehhh. Honestly, I could argue either way. My experience with replication was that it was an "all or none". If the replication process interrupts for any reason I think the whole job fails and you have zero files in that replication. But snapshots are also very fast to replicate across LAN and from 1 zpool to another. I got almost maximum line speed when I was doing Gb LAN migration.

If you have a few "ZOMG if this stuff was lost I'd cry my eyes out" I'd make a copy of those somewhere then try a snapshot/replication. If the replication fails(or the hard drive dies for good) at least you have some stuff backed up. From migrating a zpool personally nothing beats the ease of using a snapshot/replication. You let it run and you're done.. or is it "set it and forget it"? :P