[zfs-discuss] ZFS Stuck

Michael Kjörling michael at kjorling.se
Thu Dec 4 01:41:10 EST 2014

On 3 Dec 2014 21:48 -0800, from rsvancara at gmail.com (Randall Svancara):
> What I have is three, 24 disk shelves filled with 4TB disks.  The goal was
> to build out a large storage volume to hold a lot of data with redundancy
> without relying on traditional RAID cards.   In terms performance, it is
> not necessarily designed for a fast access system.  I hope to procure
> additional storage for fast access based on Luster.

In that case, dividing into 6-drive raidz2 sounds near ideal if the
resultant loss of storage to parity is acceptable. Use two drives from
each shelf for each vdev and in the best of cases (no other storage
problems) you will be able to survive the loss of an entire shelf with
no problems visible to the user. If you are able to also connect each
shelf to a separate controller and data path, as well as power supply,
that should as near as can be done eliminate single points of failure
as a possibility for outright pool failure. That best-case scenario
obviously does require that _everything_ about the other two shelves
works perfectly, however.

You have 72 disks of 4 TB each (288 TB raw capacity); with 6-drive
raidz2, that gives you 12 striped vdevs for a maximum of around 1400
uncached IOPS and a total of 192 TB usable storage, losing 96 TB to
parity. You can then later add storage in 6-disk incremenents, for
(with 4 TB drives) 16 TB worth of usable storage added per set of

Michael Kjörling • https://michael.kjorling.semichael at kjorling.se
OpenPGP B501AC6429EF4514 https://michael.kjorling.se/public-keys/pgp
                 “People who think they know everything really annoy
                 those of us who know we don’t.” (Bjarne Stroustrup)

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.

More information about the zfs-discuss mailing list