Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 28 Apr 2011 12:34:21 -0500
From:      Adam Vande More <amvandemore@gmail.com>
To:        Thomas Ronner <thomas@ronner.org>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: ZFS vs OSX Time Machine
Message-ID:  <BANLkTinSqXqM5ziFEqeiw4gnk_wtWs-u6w@mail.gmail.com>
In-Reply-To: <4DB9887A.4080109@ronner.org>
References:  <537A8F4F-A302-40F9-92DF-403388D99B4B@gsoft.com.au> <4DB9887A.4080109@ronner.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Apr 28, 2011 at 10:32 AM, Thomas Ronner <thomas@ronner.org> wrote:

> On 4/28/11 4:03 PM, Daniel O'Connor wrote:
>
>> Does anyone else use ZFS to store TM backups?
>>
>> I find that whenever my laptop (over wifi!) starts a TM the ZFS machine
>> it's backing up to grinds to a halt.. Other systems streaming stuff over NFS
>> from it also tend to stall..
>>
>
>
> Are you using zfs compression? If so, try turning that off.
>
> I have a pool with a couple of filesystems with gzip-9 compression enabled.
> Whenever I write (using zfs receive, it is a backup server) to one of those
> volumes the whole pool stalls with lots of disk activity. Even creating a
>  snapshot on another filesystem within the same pool lasts a couple of
> minutes.
>
> Does anyone know how to make this perform a little better? It's only
> writing small amounts (70-100 ops/s, 1 MB/s) on an otherwise idle pool.
> Still the drive leds blink like crazy. One of my two CPU cores is maxed out,
> the other is idle. I suppose it won't get any faster (it's CPU bound because
> of the heavy gzip compression), but why is the pool so slow? Is zfs receive
> using synchonous writes?
>

gzip-9 is very poor choice for most datasets.  It's going to be extremely
slow especially with data that can't be compressed easily eg data that
already compressed or encrypted.  So yeah, if you're running into a cpu
bottleneck, change your compression algo.  I find lzjb to be a good one for
general use.

And to the OP, I'm not familar with TM, but see if disabling sendfile in any
of your daemons helps.  Also I don't think you want this setting:
vfs.zfs.cache_flush_disable=1

-- 
Adam Vande More



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?BANLkTinSqXqM5ziFEqeiw4gnk_wtWs-u6w>