Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 19 Nov 2008 01:03:07 -0800
From:      Jeremy Chadwick <koitsu@FreeBSD.org>
To:        Nikolay Denev <ndenev@gmail.com>
Cc:        freebsd-current@freebsd.org
Subject:   Re: HEADS UP: New ZFS in the tree.
Message-ID:  <20081119090307.GA81236@icarus.home.lan>
In-Reply-To: <8ECD400F-BFE3-4E31-94F0-39AF5F44FDAC@gmail.com>
References:  <20081117205526.GC1733@garage.freebsd.pl> <8ECD400F-BFE3-4E31-94F0-39AF5F44FDAC@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Nov 19, 2008 at 10:37:32AM +0200, Nikolay Denev wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 17 Nov, 2008, at 22:55 , Pawel Jakub Dawidek wrote:
>
>> Hi.
>>
>> So ZFS was updated from version 6 to 13. Be very careful when updating
>> your system if you use ZFS. The number of changes is huge and my
>> regression tests and manual tests I did only cover part of the entire
>> functionality.
>>
>> More info here:
>>
>> 	http://svn.freebsd.org/viewvc/base?view=revision&revision=185029
>>
>> Enjoy.
>>
>> -- Pawel Jakub Dawidek                       http://www.wheel.pl
>> pjd@FreeBSD.org                           http://www.FreeBSD.org
>> FreeBSD committer                         Am I Evil? Yes, I Am!
>
> Hi Pawel,
>
> Thanks for you excellent work on ZFS!
>
> I want to report that I got again a kmem_map_too_small panic on recent - 
> - -current with the new ZFS version.
> I left the machine overnight with an endless loop running bonnie++ on a 
> raidz2 zfs pool with five disks,
> and I found it dead this morning.
> Is this still supposed to happen?
>
> I had only these two lines in my loader.conf :
>
>  vm.kmem_size="1536M"
>  vm.kmem_size_max="1536M"
>
> and i have just added vfs.zfs.arc_max="768M" and will
> run the torture test again. (the default value for arc_max was  
> 1006632960)

The ARC was set to allocate up to 1006M of the 1536M, which probably
caused kmem exhaustion.  The ARC value you picked may still be too
large, but it will require continual testing on your part.  But see
below first.

> Btw, the machine is amd64 with four gigabytes of RAM, and I have  
> upgraded the pool to version 13.

Since this is CURRENT, you should be able to increase the kmem_size
and kmem_size_max entries to something larger than 1536M, especially
since the box has 4GB of RAM.  (You shouldn't do this on RELENG_7).

You might also consider disabling prefetch; there are reports of
peoples' boxes locking up hard (requiring a hard reset) when prefetch
is enabled.  Others (like myself) just see "better overall system
responsiveness" when prefetch is disabled.

-- 
| Jeremy Chadwick                                jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20081119090307.GA81236>