Missing ZFS storage
-
I have RAIDZ2 pool of 10x 3TB disks. So I should have about 24TB usable space (or about 21.8TiB). I'm missing 1.66TiB and I can't figure out where. Here's output of zpool list:
root@storage:/ # zpool list
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
bacon 27.2T 22.7T 4.57T - 19% 83% 1.00x ONLINE -and here's zfs list:
root@storage:/ # zfs list
NAME USED AVAIL REFER MOUNTPOINT
bacon 17.3T 2.84T 17.3T /mnt/baconThis is on FreeBSD 10.2. All units above are in TiB. No snapshots, no quota, no reservations.
-
Sounds like allocation overhead.
-
That sounds right. I wonder how performance degrades if ashift is set to 9, effectively forcing 4K disks to be written as 512b.
-
That's a good question and would make an excellent lab test
-
Looks like somebody did that already:
http://louwrentius.com/zfs-performance-and-capacity-impact-of-ashift9-on-4k-sector-drives.htmlAnd apparently using large_blocks feature and 1MB record size should eliminate allocation overhead.
-
Now the real question: Can you make the adjustment without having to wipe the existing data and restore from backups?
-
@dafyre said:
Now the real question: Can you make the adjustment without having to wipe the existing data and restore from backups?
Recordsize property yes, zfs set recordsize=1M bacon, ashift no, you need to recreate the pool. I'm guessing recordsize change won't affect allocated data. So wiping and restoring is probably a valid approach.
-
@marcinozga That would be a lot of effort to backup & restore 17TB worth of stuff just to save ~1.6 TBs?
Would it be worth the effort?
-
Not really. I'll do it in a few years when it's time to upgrade disks. But most of that will end up in Amazon Cloud Drive soon.
