[GLLUG] zfs on linux - lessons learned

Jason L. Froebe jason.froebe at gmail.com
Thu Jun 12 22:31:35 EDT 2014


Hi all,

A few months back I lost several terabytes of data on a raid5 btrfs volume
causing quite a few headaches.  I switched to ZFS on Linux (non-fuse
version).  A few tidbits that might be of interest:

1) Once you create a raidz volume, the number of devices you assign it
can't be increased or decreased
2) You can not convert a mirrored volume into a raidz volume or the other
way around.
3) If the device locations change (e.g. usb3 drives), you need to export
and import the volume.
4) It is rock solid.
5) You need to disable dedup else your RAM will be consumed by ZFS.  If you
want to deduplicate your volume, expect 4GB of RAM per 1TB of storage
6) You increase the size of the raidz volume by replacing the devices with
larger devices.  Only once all the devices are the greater size will the
space become available.
7) Replacing a raidz device is done very quickly.  You'll see the old and
the new drive maxed on i/os (use nmon to view the device utilization easily)
8) Compression on subvolumes (zfs: file systems) works seamlessly.  You set
it and forget it.
9) Use the deadline or noop i/o scheduler to maximize your i/o throughput
and minimize latency.

Has anyone else used ZFS on Linux?

-- 
To stay young requires unceasing cultivation of the ability to unlearn old
falsehoods. -- Lazarus Long, "Time Enough For Love"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.egr.msu.edu/mailman/public/linux-user/attachments/20140612/9e1b3124/attachment.html>


More information about the linux-user mailing list