Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 03 Jan 2009 09:34:24 -0200
From:      Sergio de Almeida Lenzi <lenzi@k1.com.br>
To:        freebsd-questions <questions@freebsd.org>
Subject:   Re: Using HDD's for ZFS: 'desktop' vs 'raid / enterprise' -edition drives?
Message-ID:  <1230982465.7117.29.camel@lenzix.cwb.casa>
In-Reply-To: <495F2919.6040103@isafeelin.org>
References:  <495E17AD.30707@isafeelin.org> <20090102160727.A38841@wojtek.tensor.gdynia.pl> <F881B4D0-69A3-4E33-BA55-EC5947064467@gmail.com> <495F2919.6040103@isafeelin.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Hello....

About your zfs experience...  please let me tell mine..

I was having toubles with ufs2 on some small servers (desktop grade)
machines running postgresql (64 bits, or 32 bits) all running FreeBSD
7.X
some still 6.X...
the problem is that sometimes the database brokes because the ufs2 fsck
wipes out the pg_log files (that holds metadata, and so the database is
lost/compromiesed)

it used to be aobout 1 or 2 times a week, well you willl say that it is
very often... 
but there are more thatn 1200 servers... that is about a database
problem in 16 years
the machines are in remote zones and cannot stop..

So a month ago I decided to try zfs... first on onpensolaris (that, as
expected, works very well...)
than in a "set"  of 4 machines running FreeBSD 7.0/7.1 (a small
partition to boot, than filesystem / on zfs).
1 of them with 2gb of memory, the others with 1g, 512m, 256m (I know
that zfs is unstable with less than 512mb...)
but it is only for testing.... the machines varies from amd64 to
celeron... (64 and 32 bits..) NO RAID
on the small machine, a buildworld lasted 2 days...

All running a database (test of course) without no break. with a custom
application that
updates  tables (several thousand rows, with foreing keys and
triggers...)  and then rollback...
the drives area always with access light on... 

Sometimes (several times a day, random..) the machines are switched
off..  without shutdown...
and than switched on again...  about (5 -10 times) that is about 60 * 8
-> 480 power on/power off cycles each machine

Well. I can say that I still have not lost a database... it is
incredible fast, reliable
comes up without any fsck wait time. in less than a minute... 

Now I will start to put it on Dells and those PERC controllers... 

That is my experience...

Sergio



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1230982465.7117.29.camel>