RAID6 URE Risk Acceptable, or Have to Go RAID10?
My client has a 12-bay Synology DiskStation DS3617xs NAS filled with 14TB Seagate IronWolf Pros. It will be used in an adjacent building with Active Backup for Business to backup the server and 40 workstations.
Sadly, according to support, triple parity isn't an option with Synology, so I'm trying to decide whether it's reasonably safe to go with RAID6 or if the only logical choice is RAID10.
I've understood for a while that the size of modern hard drives makes RAID5 a bad idea due to the likelihood of an unrecoverable read error during a rebuild. I've seen a small amount of discussion on whether RAID6 is a good idea, but haven't come across any details other than rebuild time. I have no first-hand experience for what kind of duration to expect when rebuilding a 140TB array.
My own take on whether RAID6 is reasonably safe is that a rebuild would only fail with a single drive failure if two of the remaining drives both had UREs on sectors that contain the exact same data. Granted, if you have two concurrent drive failures, you're in the same boat as with one drive failure on RAID5, nervously biting your nails and praying that the rebuild finishes.
Is my understanding of the minimal risk involved with UREs and RAID6 accurate? How much time is it likely to take to rebuild after a drive failure?
Thanks!
Sadly, according to support, triple parity isn't an option with Synology, so I'm trying to decide whether it's reasonably safe to go with RAID6 or if the only logical choice is RAID10.
I've understood for a while that the size of modern hard drives makes RAID5 a bad idea due to the likelihood of an unrecoverable read error during a rebuild. I've seen a small amount of discussion on whether RAID6 is a good idea, but haven't come across any details other than rebuild time. I have no first-hand experience for what kind of duration to expect when rebuilding a 140TB array.
My own take on whether RAID6 is reasonably safe is that a rebuild would only fail with a single drive failure if two of the remaining drives both had UREs on sectors that contain the exact same data. Granted, if you have two concurrent drive failures, you're in the same boat as with one drive failure on RAID5, nervously biting your nails and praying that the rebuild finishes.
Is my understanding of the minimal risk involved with UREs and RAID6 accurate? How much time is it likely to take to rebuild after a drive failure?
Thanks!
If a drive fails and another has UREs on it then RAID 6 can cope, RAID 10 won't cope though if the URE is on the mirror of the failed one.
ASKER
Good info. Thank you.
Still, my most pressing question is regarding a single drive failure in RAID6 and the risk UREs pose in this configuration
If, during the rebuild, two drives each encounter a URE, does that only cause the rebuild to fail if the UREs occur on the exact same bits of data on both drives (a highly unlikely scenario) or will it fail simply because two drives encountered a URE, regardless of whether it was on the same bits of data (likely enough to warrant consideration)?
Still, my most pressing question is regarding a single drive failure in RAID6 and the risk UREs pose in this configuration
If, during the rebuild, two drives each encounter a URE, does that only cause the rebuild to fail if the UREs occur on the exact same bits of data on both drives (a highly unlikely scenario) or will it fail simply because two drives encountered a URE, regardless of whether it was on the same bits of data (likely enough to warrant consideration)?
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
Thank you for your assistance!
RAID-6 with 2 drives failing = RAID-5 with one failing drive: All disks need to be read to calculate the data / parity records. (for ALL data).
(Failure is the same for ALL disks)
RAID-10 with one disk failing means one of the RAID-1 sets in the striped array needs to be rebuilt. (straight one disk copy ).
With 2 disks failing it depends.... if the disks fail in the same mirrorset the RAID-10 dies irrecoverably.
if disks fail in different mirrorsets then recovery time will be straight diskcopies. (speed depending on bandwidth).