One of my drives failed over-night. In the SMART stats, "Current pending sector" and "Offline uncorrectable" seem to be new issues while the "UDMA CRC error count" occurred a week ago and didn't disable the disk (and I replaced the cable upon seeing it).

 

I plan on doing a SMART extended test. I'm in the process of rebuilding another failed disk, so I'll wait for that finish. Luckily, I just finished building a second parity a couple days ago.

 

Does unRAID prioritize against writing to a disabled/emulated disk? And is there a way to see if it was a write or read that brought it down (or does unRAID only disable on write)? Trying to figure out if I can skip rebuild.

 

Edited December 25, 20205 yr by golli53

39 minutes ago, golli53 said:

Does unRAID prioritize against writing to a disabled/emulated disk? And is there a way to see if it was a write or read that brought it down (or does unRAID only disable on write)? Trying to figure out if I can skip rebuild.

No, the emulated disk is used normally. Only a write will disable a disk, however, a failed read will trigger a write, as unraid will attempt to rewrite the parity calculated content when a read fails.

 

Skipping the rebuild won't save you any time, as you would need to do a correcting parity check to bring parity back in line with the disk content. And, if anything was changed on the emulated disk after the physical disk was dropped, those changes will be lost.

On 7/24/2019 at 9:21 AM, jonathanm said:

No, the emulated disk is used normally. Only a write will disable a disk, however, a failed read will trigger a write, as unraid will attempt to rewrite the parity calculated content when a read fails.

 

Skipping the rebuild won't save you any time, as you would need to do a correcting parity check to bring parity back in line with the disk content. And, if anything was changed on the emulated disk after the physical disk was dropped, those changes will be lost.

Thanks. I ended up replacing it. In general, wish it would prefer user share writes away from the disk to keep parity valid, although I guess it might be impossible to tell if the first write error that brought the disk down actually affected the data or not.

2 minutes ago, golli53 said:

wish it would prefer user share writes away from the disk to keep parity valid,

If it's important to you, create a feature request to add a configuration setting that would immediately force the file system on an emulated disabled disk to remount read only. It would need to be an optional setting, as most folks would rather be able to keep using their server normally while they rebuild.

 

I can see the value in certain circumstances though, so if you make a feature request you may get some traction, who knows.