Hey all,

 

so a week or so ago, I found some issues with sectors on various drives. I ended up buying 4 new 3tb reds so far I have rebuilt disk 5, disk 6 and finally the parity drive (46 errors during parity sync no problems with this drive prior) thankfully I lost no data in this adventure, and those drives were bad after pre-clearing each of them the pending sectors stayed and off line errors, multi-zone errors and reallocated sectors were evident, but my disk 1 is still in the server and it shows 65535 pending sectors with no other errors and I have no rebuilt two data drives and the parity drive, did a parity sync and two parity checks and still no errors on disk 1 other then those pending sectors. I am not going to take any chances, I am still going to replace it and pre-clear to check and rma if required but is there really a problem? No other errors, nothing in the log, I did have some read errors when this all started but I replaced the PSU and nothing since. One other bad drive showed 65534 pending sectors prior to replacement and eventually had 120 reallocated sectors after the pre-clear was done.

 

thanks,

 

Dave

http://lime-technology.com/wiki/index.php/Troubleshooting#Resolving_a_Pending_Sector

yes read that several times, thanks. My question is after rebuilding data on 2 new 3 TB drives and then replacing a parity drive and having that sync and parity check twice and no further errors on the drive in question - Is there really any problems with it.

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it? I have never had any pending sectors yet so I have no real world experience but that is what I have read. The real question is even if the drive with 65535 pending sectors was "OK", do you really want to trust your data living on that drive?

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it?

Nope. Unraid does not red ball a drive unless a write to it fails. It doesn't do any SMART checking.

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it?

Incorrect.

I have never had any pending sectors yet so I have no real world experience but that is what I have read. The real question is even if the drive with 65535 pending sectors was "OK", do you really want to trust your data living on that drive?

It would show"errors" if those sectors were found to be un-readable in a parity check, but no other symptoms.

 

No drive with any pending sectors is OK. (you cannot use it to re-construct the same sectors on any other disk in the array if they were to fail.)

 

Honestly, I would RMA the drive, even if it looked fine, as your error sounds more like a firmware issue than an actual disk failure.

You can try to remove it from the array and preclear it (basically giving the drive the chance to re-allocate those pending sectors)

Then see what the smart report says.

 

You would see a RED indicator, and the drive disabled (and emulated ) if ANY write failed to the drive.

( just 1 failed write would disable the drive )

 

Joe L.

 

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it?

Nope. Unraid does not red ball a drive unless a write to it fails. It doesn't do any SMART checking.

 

OK understood. Just to make sure I get this I want to ask a question using an example. Lets say a drive has 20 pending sectors, unRAID does not monitor those pending sectors so everything will appear to be fine. I happen to copy something to unRAID and it attempts to write to one of the pending sectors and fails and therefore the drive "red balls"?

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it?

Nope. Unraid does not red ball a drive unless a write to it fails. It doesn't do any SMART checking.

 

OK understood. Just to make sure I get this I want to ask a question using an example. Lets say a drive has 20 pending sectors, unRAID does not monitor those pending sectors so everything will appear to be fine. I happen to copy something to unRAID and it attempts to write to one of the pending sectors and fails and therefore the drive "red balls"?

Only if the drive actually fails the write. A pending sector isn't the end of the world for a drive, they come from the factory with LOTS of spare sectors, and in normal operation, if a sector takes too long to read, the data will be written instead to one of the spares, and the OS is none the wiser. A current pending sector is one that the drive can't read the data from. If a drive finally gives up trying to read the sector and remap it for itself and spits out a read error, unraid immediately spins up all the other drives, does the parity calculation to figure out what was SUPPOSED to be in that sector, and writes it back to the drive. Unraid also increments the error column on that drive, so you have an idea that SOMETHING went wrong, but unraid was able to successfully handle it. An otherwise healthy drive will clear the pending sector counter for that spot, and increment the reallocated sector count. If the WRITE to the drive fails, unraid gives up trying to use the drive, red balls it, and all future operations for that drive are actually being done on the parity drive, and the red ball drive is calculated on the fly from all your drives.

 

The reason a pending sector is death to a rebuild, is that during a rebuild, unraid is using the data from that sector to determine what is supposed to be on another drive. It can't calculate what's supposed to be there and write it back. So, a pending sector will cause a corrupt drive rebuild. Regular parity checks are a good way to make sure all drives can be successfully read end to end so a rebuild will succeed.

Thanks jonathanm, very good explanation. I just bookmarked your post for the future.

My understanding was that if you had a drive with any pending sectors the unRAID server would show a red ball next to it?

Nope. Unraid does not red ball a drive unless a write to it fails. It doesn't do any SMART checking.

 

OK understood. Just to make sure I get this I want to ask a question using an example. Lets say a drive has 20 pending sectors, unRAID does not monitor those pending sectors so everything will appear to be fine. I happen to copy something to unRAID and it attempts to write to one of the pending sectors and fails and therefore the drive "red balls"?

Only if the drive actually fails the write. A pending sector isn't the end of the world for a drive, they come from the factory with LOTS of spare sectors, and in normal operation, if a sector takes too long to read, the data will be written instead to one of the spares, and the OS is none the wiser. A current pending sector is one that the drive can't read the data from. If a drive finally gives up trying to read the sector and remap it for itself and spits out a read error, unraid immediately spins up all the other drives, does the parity calculation to figure out what was SUPPOSED to be in that sector, and writes it back to the drive. Unraid also increments the error column on that drive, so you have an idea that SOMETHING went wrong, but unraid was able to successfully handle it. An otherwise healthy drive will clear the pending sector counter for that spot, and increment the reallocated sector count. If the WRITE to the drive fails, unraid gives up trying to use the drive, red balls it, and all future operations for that drive are actually being done on the parity drive, and the red ball drive is calculated on the fly from all your drives.

 

The reason a pending sector is death to a rebuild, is that during a rebuild, unraid is using the data from that sector to determine what is supposed to be on another drive. It can't calculate what's supposed to be there and write it back. So, a pending sector will cause a corrupt drive rebuild. Regular parity checks are a good way to make sure all drives can be successfully read end to end so a rebuild will succeed.

 

This needs to be stickied and Wiki'd and all sorts of saved! Excellent write up, Jonathan! Thank you very much.

This needs to be stickied and Wiki'd and all sorts of saved! Excellent write up, Jonathan! Thank you very much.

 

Agreed! It would be nice if the more technical details of how unRAID worked would be easily accessible on a wiki or on lime-tech's website. I have to imagine a lot of unRAID users would find that pretty useful as most users seem pretty computer savy or at least willing to learn.