RAID5 requires at least 3 disks and RAID6 requires at least 4. But Rockstor’s Data Loss Prevention page mentions having an extra drive http://rockstor.com/docs/data_loss.html#data-loss-prevention-and-recovery-in-raid5-pools . But why is that? If a RAID5 has, by definition, 1 disk failure protection – what does having another disk do to help prevent data loss?
I do not see any mentions of extra disk in the linked page, but usually when an extra disk is mentioned during RAID conversations it means having a HOT spare.
i.e. having extra drive in the chassis to be available if you loose a drive in raid set.
this way you can add the drive to the array and start rebuilding the array as soon as you know there is an issue. you can get a new drive and replace the failed one with it as a hot spare. this does several things.
first , it helps to minimize time you run the array in degraded mode with a failed drive. second, it speeds up the time from degraded to rebuild to back to normal thus minimizing the chances of second drive failure occurring while you are unprotected. also with BTRFS setup, it helps you to manage the array size, providing you have space in the chassis, by expanding the array at any time and adding a new hot spare at your leisure.
“Currently, raid5 is experimental and we suggest that you don’t create a pool
with the minimum configuration of 3 drives. It’s very error prone to replace a
failed drive of a 3 drive raid5 pool.”
btrfs has issues recovering when the number of drives falls below the minimum number of disks due to a failure of a disk. Sometimes it works sometimes it doesn’t definitely better than it used to though.
missed that. LOL
I guess my answer is not good after all.
it does apply to a normal RAID 5/6 setup but not BTRFS pool.
Spectre694 has it right.
I guess my biggest issue with Rockstor is that it will not allow the use of partitioned drives in the pool. not really sure how it is now, but roughly 6 month ago BTRFS had issues with any raid level pools on raw devices.
you could loose the pool with all the data if one drive fail even in raid-1 (been there, done that)
but on raid pools build using partitions it was fine.
right now I only build BTRFS pools on partitioned drives for my file server.