Marking a disk as low priority?

[jonatkins]

This is probably a daft question, but I'll ask it anyway...

Is there a way of marking a disk as low priority? Specifically, marking it to only be read from when absolutely required - i.e. when data cannot be satisfied through redundancy in the rest of the pool.

Why? I'm working on transferring data from an old raidz1 pool to something with more redundancy, and one of the disks has patches of bad sectors.

The options I see are:
- leave the bad disk in place, and put up with periods of awful read performance as it hits bad sectors.
- remove the disk from the pool, and pray real hard there's not a single bad sector on any of the rest of the disks.

Is there a third way? Avoid reading from the disk when ever possible, but leaving it in place for emergency use if another disk in the pool has issues?

Jon

 

[Bidule0hm]

No, you can't AFAIK.

Whatever you do, do not do that: "remove the disk from the pool, and pray real hard there's not a single bad sector on any of the rest of the disks." if you care about your data.

Your options are:
- replace the bad drive with a new drive
- accept the low perf on bad sectors

 

[garm]

Just let ZFS do it’s thing and get the data off the pool asap or replace the drive

Edit: ninjad

 

[jonatkins]

Wouldn't "replacing the drive" also be "praying the rest of the disks have no bad sectors", as there'd be no redundancy during the resilver?

Would replacing the drive with the bad disk still online preserve the redundancy during the resilver? (Although that's not an option here, as the old machine has no free SATA connections for this)

To clarify, I *am* in the process of pulling the data off, onto a new machine with a raid-z3 pool. It's just taking an awful long time when it hits the bad bits...

 

[Bidule0hm]

Wouldn't "replacing the drive" also be "praying the rest of the disks have no bad sectors", as there'd be no redundancy during the resilver?

No because exactly this:

Would replacing the drive with the bad disk still online preserve the redundancy during the resilver?

To clarify, I *am* in the process of pulling the data off, onto a new machine with a raid-z3 pool. It's just taking an awful long time when it hits the bad bits...

Well, just wait, that's what I'd do, I think it's the best option here.

 

[garm]

Don’t offline the drive, simply replace it. There shouldn’t be any writes during resilvering and if it can’t resilver you won’t be able to backup the data any way