0

About hdd URE, I knew these points:

  1. For some reasons, when harddisk reading a sector that the FEC(Foward Error Correction data) could not correct the errors on that sector, we encontered an URE.
  2. The rating we encountered an URE is very low, but still exists.
  3. When reconstructing a RAID 5 array, sometimes it will happened and the reconstruction progress will stop.

But I still have some questions:

  1. If there is a single disk, what will happened? Hardware/file system report an error and we lost a file? Or we got the file with wrong data?
  2. Will rewrite some data to that URE sector could turn the sector become normal? Or must we use some utilities provieded by HDD manufactory and remap another reserve sector?
  3. If it happened when we mirror/re-mirror a RAID 1/10 array, what will the RAID controller do? Stop the mirror progress? Or just copy the uncorrect data to another disk?

Thanks for the answer, question 1&2 is solved.

But the 3rd question I mean if encountered URE when converting a single HDD to RAID 1 array by add another new disk, or replacing a failure disk in a RAID 1/10 array, there's no redundancy to correct the error. Will it complete the mirror/re-mirror progress with error data? OR stop the progress like RAID 5 recontruction?

Harley
  • 11
  • 4

1 Answers1

1
  1. With a single disk, an unrecoverable error is just that - it can't be done and it is reported to the filesystem and subsequently to the application trying to read a file. Generally, it is preferred to get an explicit error instead of unreliable data.
  2. Writing to an unreadable sector will either fix the physical sector (for a soft error when e.g. writing was interrupted by power loss) or the drive will map the logical sector to one of its spare pool. This happens at the discretion of the drive and isn't usually user/driver selectable.
  3. The RAID controller will most likely repair the sector - either from the mirror or by rebuilding the data from the redundancy set. When during a mirror or rebuild a(nother) read error prevents this repair, the error sticks and the array is bad. Some RAID sets can repair multiple errors (RAID 6 or some nested RAIDs) but once errors pile up you're out of luck.

It's important to make sure errors don't pile up on rarely used sectors - they can become uncorrectable errors when sectors are left unread for months or even years. So, make sure you enable data scrubbing, media patrol, patrol read or whatever it's called on your hardware to check all data on a regular basis. That way, you ensure a rebuild works when you need it.

Some people report that during a rebuild, additional drives start to fail because of the stress but I've found this to be a myth. The drives just stumble on stale, accumulated errors. You can stress even very old drives for days without any problems.

Zac67
  • 10,320
  • 2
  • 12
  • 32
  • Thanks for the answer, but the answer for Q3 is not what I asked. I edit the question, could you answer Q3 again? – Harley Dec 27 '17 at 09:39
  • I answered it in a more general way - check the 2nd sentence, 'another read error' assuming that the mirror/remirror/rebuild was triggered by a fault. A disk will not let you read a defective, physical sector and if the controller can't substitute or reconstruct you can't read the logical sector. – Zac67 Dec 27 '17 at 09:43
  • So the progress will stop? For a RAID 1/10 build by 8TB disk and full of data, that's a high chance encountered URE when re-mirror about 47% for 10e-14 disks or 6.2% for 10e-15 disks? – Harley Dec 27 '17 at 11:24
  • Since the process can't continue it'll (usually) stop. Some controllers might choose to mirror/rebuild the rest but that's not common. The general assumption is to restore from backup to a freshly initialized array. – Zac67 Dec 27 '17 at 11:55
  • So, if I need a RAID with 10TB capacity with redundancy, and it should works 7x24, I'd better choose RAID 6? And backup to a RAID 1/5/10 volume? – Harley Dec 28 '17 at 01:27
  • There's no simple answer to that. Choosing the RAID level and the type of HDDs/SDDs to use is a trade-off between reliability, performance (rebuild time) and cost. More drives are faster than fewer, RAID 6 is more reliable than RAID 5 and so on. – Zac67 Dec 28 '17 at 11:10
  • How you do your backup should be asked in another question. – Zac67 Dec 28 '17 at 11:23