Linked by Thom Holwerda on Thu 12th Jan 2012 22:54 UTC
FreeBSD Some people already submitted this news last week, but it wasn't until today that it became official: the FreeBSD team has announced the release of FreeBSD version 9.0. As you may expect from the major version number change, this is releas eis packed with new stuff.
Thread beginning with comment 503153
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE: Time to give it a try
by 2shanfernando on Fri 13th Jan 2012 09:38 UTC in reply to "Time to give it a try"
2shanfernando
Member since:
2008-02-01

Same, its got ZFS version 28 support which caught my eye... I'm still running OpenSolaris SNV_134 which has ZFS 22.

Anyone experienced with ZFS on FreeBSD want to comment on its stability?

Reply Parent Score: 2

RE[2]: Time to give it a try
by TTy23 on Fri 13th Jan 2012 12:20 in reply to "RE: Time to give it a try"
TTy23 Member since:
2010-05-28

I used it for quite some time now, and had no issues at all. Which does not mean that there are any, if you want to be sure, take a look at the mailing list archives: http://lists.freebsd.org/pipermail/freebsd-fs/

Unfortunately ZFS version 28 is the last one released by Oracle to the public. Later versions will not be available, so the good stuff like encryption will probably not come to FreeBSD for quite some time.

Reply Parent Score: 2

RE[3]: Time to give it a try
by Kebabbert on Fri 13th Jan 2012 13:59 in reply to "RE[2]: Time to give it a try"
Kebabbert Member since:
2007-07-27

Unfortunately ZFS version 28 is the last one released by Oracle to the public. Later versions will not be available, so the good stuff like encryption will probably not come to FreeBSD for quite some time.

Lot of original ZFS developers has joined Illumos/OpenIndiana. Even the head architect and creator of ZFS have quit Oracle. These developers from Sun have already developed ZFS functionality that Oracle does not have. The FreeBSD developers does not know how to develop ZFS functionality, they just port it from the original ZFS developers.

So if you run OpenSolaris, you can easily upgrade to OpenIndiana via IPS package systems, and then you can dualboot via snapshots in GRUB.

Reply Parent Score: 2

RE[2]: Time to give it a try
by Laurence on Fri 13th Jan 2012 15:38 in reply to "RE: Time to give it a try"
Laurence Member since:
2007-03-26

Same, its got ZFS version 28 support which caught my eye... I'm still running OpenSolaris SNV_134 which has ZFS 22.

Anyone experienced with ZFS on FreeBSD want to comment on its stability?

I've been running FreeBSD 8.1 (I think) for about a year and ZFS has been rock solid.

I've also run ZFS on Nexenta and OpenSolaris, and quite honestly, I prefer FreeBSD as I've not experienced any performance issue with ZFS but also I get the added bonus of a more preferable (for me) OS.

I'd definitely recommend it.

Reply Parent Score: 3

RE[2]: Time to give it a try
by phoenix on Fri 13th Jan 2012 19:30 in reply to "RE: Time to give it a try"
phoenix Member since:
2005-07-11

We've been using ZFS since it hit 7-STABLE way back when. Other than some growing pains around memory usage and poor vdev setup (don't do a single 24-disk raidz2!!), things have been very stable.

We just put into production 2 FreeBSD 9.0 storage boxes (one upgraded from 8.2, the other clean install of 9.0) using raidz2, lzjb compression, and dedupe. These are backups boxes running our custom rsync scripts to backup over 150 servers every night, creating snapshots every morning, and pushing the data to an offsite server during the day.

The things to note for using ZFS on FreeBSD:
- use 64-bit hardware and version of FreeBSD
- stick as much RAM as you possibly can into it
- use the fastest drives you can afford (avoid Green drives)
- be sure to stick an SSD in the system for use as L2ARC (read cache)
- if you can afford a fast SSD, use it for SLOG (write cache)

If you are going to play with dedupe, you need at least 16 GB of RAM, preferably more. And your RAM needs will only increase as you add data to the pool (the rough calculation is about 1 GB of RAM per TB of deduped data in the pool). You can use an L2ARC device to lower the RAM requirements, but that only helps so much.

If you want performance, go with mirror vdevs. If you want storage space, go with raidz2 (or even raidz3) vdevs with small numbers of drives per vdev (keep it under 10). The more vdevs in the pool, the better the overall performance of the pool (meaning a 24-disk pool using 1 raidz2 vdev will be much slower than a 24-disk pool with 4x 6-drive raidz2 vdevs).

It may take a little tuning of /boot/loader.conf settings to make things perfectly stable, although the auto-tuning in 9.0 has improved a lot. All we set now is the vfs.zfs.arc_max to about 80% of RAM.

Reply Parent Score: 3