Personally, this is what I watch.

There are other problems that aren't DISK error or FILE SYSTEM problems that can arise, such as the mover not moving.

But those show up as 'disk full' sort of messages in your PC when you try to copy stuff to unRAID.

http://lime-technology.com/forum/index.php?topic=1462.msg10347#msg10347

 

This does not explain what happens if a drive has a read error during the rebuild of a disk.

 

Maxse, don't ignore the pending sectors. Parity based protection relies on the systems ability to read every other drive in the event of a disk failure. Pending sectors are unreadable areas of the drive. They will prevent the correct rebuilding of a failed drive.  See the link I gave earlier.

The errors that you see on unRAID don't tell you everything.  For instance a drive can only remap so many sectors, and then it runs out of sectors it can remap, and when that happens you're in trouble.  Any sectors that should be re-mapped now cannot be.  Smart tests should make you aware of issues like this.

Thanks for explaining it. Wow.

But every time there is a smart error it will show up as a general error in the column wouldn't it? Like what happened to me basically?

 

Otherwise what's the point of having a red ball drive when in reality there could already be errors that will prevent a successful rebuild. Am I understanding it right?

 

This makes the main errors page completely useless and false sense of security. Doesn't it?

 

 

Sent from my iPhone using Tapatalk

Guys how does Synology deal with all this by the way?

 

I'm beginning to understand that main page with all the errors is completely useless? Because when a drive red balls , if you didn't run the smart testing script to check you could already have catastrophic errors that won't let you recover??

 

 

Sent from my iPhone using Tapatalk

http://lime-technology.com/forum/index.php?topic=1462.msg10347#msg10347

 

This does not explain what happens if a drive has a read error during the rebuild of a disk.

 

Maxse, don't ignore the pending sectors. Parity based protection relies on the systems ability to read every other drive in the event of a disk failure. Pending sectors are unreadable areas of the drive. They will prevent the correct rebuilding of a failed drive.  See the link I gave earlier.

 

The only reason I am pushing back is I don't want people to think it is not possible to rebuild a disk if there is a disk in the array with a pending sector. Although you are right and in some circumstances it COULD result in the drive returning a read error, this is not the common scenario. I have seen many many many many many parity checks that run perfectly with no error counts increasing and no indication of read errors in the log, while pending sectors and even reallocated sectors are increasing. And if the parity check runs perfectly the rebuild would too.

 

I once asked a question about what would happen if there was a read error REPORTED BACK TO THE OS during a rebuild. I can't find the link but I believe Tom said it would not fail the rebuild but that the error count would increment. So minimal corruption could result while still not stopping the disk from being rebuilt (this is what we would want IMO).

 

Guys how does Synology deal with all this by the way?

 

I'm beginning to understand that main page with all the errors is completely useless? Because when a drive red balls , if you didn't run the smart testing script to check you could already have catastrophic errors that won't let you recover??

 

It is important to monitor the smart attributes. If you run myMain there is a single screen that will show you all of the smart issues on all of your drives on a single page. It gives some cool (if I do say so myself) features to suppress issues you've already researched and accepted, that will only pop up again if the value gets worse. I suggest using that after each parity check to see if you are starting to develop any SMART issues. Under the screen is documentation on all of the smart attributes.

 

You can click on the myMain instructions link in my sig and scroll down a little to see a screenshot and instructions for using the smart view.

Conceptually I agree with your post from a few years ago r.e. the process that SHOULD happen when a read error occurs -- i.e. UnRAID gets notified there's a read error; so it corrects the data by reading all the other disks, then writes the data back to the offending disk ... which should write it to a newly re-mapped sector.   

 

The fact that the data can be corrected is why UnRAID doesn't red-ball a disk based on READ errors ... only on write errors.

 

The flaw in this logic, however, is that the sector should no longer be "Pending" in the SMART status ... if it's truly been re-mapped, then it should now be a re-mapped sector -- NOT a pending sector.    Makes me wonder if UnRAID is only correcting the data when this happens ... but not actually writing it back to the disk.

 

Tom ?? ... it'd be nice to get some feedback on the exact mechanism UnRAID uses in this circumstance.

 

 

 

Tom did confirm later in that thread.

http://lime-technology.com/forum/index.php?topic=1462.msg10347#msg10347

 

This does not explain what happens if a drive has a read error during the rebuild of a disk.

 

Maxse, don't ignore the pending sectors. Parity based protection relies on the systems ability to read every other drive in the event of a disk failure. Pending sectors are unreadable areas of the drive. They will prevent the correct rebuilding of a failed drive.  See the link I gave earlier.

 

The only reason I am pushing back is I don't want people to think it is not possible to rebuild a disk if there is a disk in the array with a pending sector. Although you are right and in some circumstances it COULD result in the drive returning a read error, this is not the common scenario. I have seen many many many many many parity checks that run perfectly with no error counts increasing and no indication of read errors in the log, while pending sectors and even reallocated sectors are increasing. And if the parity check runs perfectly the rebuild would too.

 

I once asked a question about what would happen if there was a read error REPORTED BACK TO THE OS during a rebuild. I can't find the link but I believe Tom said it would not fail the rebuild but that the error count would increment. So minimal corruption could result while still not stopping the disk from being rebuilt (this is what we would want IMO).

 

Guys how does Synology deal with all this by the way?

 

I'm beginning to understand that main page with all the errors is completely useless? Because when a drive red balls , if you didn't run the smart testing script to check you could already have catastrophic errors that won't let you recover??

 

It is important to monitor the smart attributes. If you run myMain there is a single screen that will show you all of the smart issues on all of your drives on a single page. It gives some cool (if I do say so myself) features to suppress issues you've already researched and accepted, that will only pop up again if the value gets worse. I suggest using that after each parity check to see if you are starting to develop any SMART issues. Under the screen is documentation on all of the smart attributes.

 

You can click on the myMain instructions link in my sig and scroll down a little to see a screenshot and instructions for using the smart view.

 

Agreed, pending sectors don't stop a rebuild and they shouldn't. However, the severity of the issue is subject to debate without further input from Tom. I am assuming the worst without further clarification.

Conceptually I agree with your post from a few years ago r.e. the process that SHOULD happen when a read error occurs -- i.e. UnRAID gets notified there's a read error; so it corrects the data by reading all the other disks, then writes the data back to the offending disk ... which should write it to a newly re-mapped sector.   

 

The fact that the data can be corrected is why UnRAID doesn't red-ball a disk based on READ errors ... only on write errors.

 

The flaw in this logic, however, is that the sector should no longer be "Pending" in the SMART status ... if it's truly been re-mapped, then it should now be a re-mapped sector -- NOT a pending sector.    Makes me wonder if UnRAID is only correcting the data when this happens ... but not actually writing it back to the disk.

 

Tom ?? ... it'd be nice to get some feedback on the exact mechanism UnRAID uses in this circumstance.

 

I agree. If this is happening then any pending sector would be resolved immediately and none would ever appear in SMART.