I noticed recently that two of the drives in my array (v5.05) were showing error counts in the GUI (I was prepping for a migration to v6).  I ran SMART reports on the offending drives, and both drives are showing re-allocated sectors, pending sectors for re-allocation and entries in the SMART error log (attached).  Both drives appear to be responsive, but based on the SMART data, are either partially failing or going to fail very soon.

 

What's the best strategy to maximize my chance of no data loss?  Run a parity check on the array as is and hope those sectors get properly reallocated on at least one of the drives (and then replace the drive that couldn't be helped), or replace one of the drives first (probably sde based on the higher number of sectors pending allocation) and then run a parity check?

smart.sde.txt

smart.sdo.txt

I wouldn't run a correcting check, there's a chance of corrupting parity, you could run a non correcting check but don't think that would help with the pending sectors.

 

I would replace the worst disk first, if there are read errors from the other during the rebuild then use a file compare utility between the readable files of the old disk (should be most of them) and the rebuilt one, assuming that you don't have checksums.

Thanks Johnnie, I figured as much.  So in the last few days I ran a preclear on a 8TB Seagate archive drive and started the parity rebuild by swapping out sde.  Parity finished but as I feared, I'm seeing that sdo is showing 28 errors in the error column in the unRAID web GUI home page.  So when the parity rebuild runs into a read error on a specific disk, what exactly happens?

There will be some corrupt file(s) on the rebuilt disk, 28 errors are very few, with some luck and if you have mostly media files it will be a small flicker during a movie or tv show.