2 failing drives, best way to switch things over?

[cmer]

I have an array of 16x3TB drives, including 2 parity.

 

I currently have 2 drives giving me a SMART error. 1 parity and 1 data. The exact error is: "End-to-End Error, Failing now" with a count of 1 in Attributes.

 

First of all, is it fair to assume that I should decommission those drives now?

 

Second, what's the best way to do that? Many of my data drives are still empty, so I'd like to move the data from the failing data drive to an empty drive, and hopefully move the parity drive to another disk without rebuilding parity.

 

Here are my assumptions. Maybe I'm wrong.

 

- I can just `mv` files from /mnt/disk5 (failing) to /mnt/disk10 (new, empty) then remove disk5 from my array.

- There is no way to move my parity drive without rebuilding parity. I'd have to decommission my failing parity drive, then assign parity to a new drive, then rebuild. I'm hoping for a way to avoid rebuild!

 

What'd be the best way to proceed?

Thanks!

Edited November 2, 2017 by cmer

[JorgeB]

38 minutes ago, cmer said:

First of all, is it fair to assume that I should decommission those drives now?

 

Yes

 

38 minutes ago, cmer said:

move the parity drive to another disk without rebuilding parity.

 

You can't do that.

 

If you can keep the server data unchanged during the rebuild I would replace both disks at the same time.

[trurl]

^^^What johnnie said.

 

Also, I don't generally recommend moving files from failing drives to other drives in the array. That just causes a lot of disk activity while the array is compromised. Moving means writing to the destination, which will also write parity, and then deleting from source. And the delete is actually a write operation on the failing drive, and that write also writes parity.

 

If you only had one parity and it was failing along with a data drive, then I might recommend instead copying the data from the failing data drive to another system or to an unassigned device outside the array. That would only involve reading an array drive and no writes to the compromised array.

 

 

[cmer]

Thanks guys, really appreciate the help!