Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 18 Apr 2008 13:15:38 -0500
From:      "James Snyder" <jbsnyder@gmail.com>
To:        "Claus Guttesen" <kometen@gmail.com>
Cc:        freebsd-stable@freebsd.org
Subject:   Re: ZFS & Bittorent -> Hang?
Message-ID:  <33644d3c0804181115v38611102yc48c2e46f30d0bf6@mail.gmail.com>
In-Reply-To: <b41c75520804181103p1719810dsdf22a7c151e9f30e@mail.gmail.com>
References:  <16447331.post@talk.nabble.com> <ft2fsj$7i7$1@ger.gmane.org> <16491496.post@talk.nabble.com> <b41c75520804150244p1f7c9693w180cb3a0d4416e4c@mail.gmail.com> <33644d3c0804151005r17095cb1n33117a8e4d8cc09b@mail.gmail.com> <16763410.post@talk.nabble.com> <b41c75520804181103p1719810dsdf22a7c151e9f30e@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
No problem.  I've filed a PR:
http://www.freebsd.org/cgi/query-pr.cgi?pr=122888

This doesn't seem to be a volume-of-data thing, not sure if it is
related to using raidz or not.  I don't have another spare disk in the
machine I tested on to pare things down.

Otherwise, seems to be pretty stable.  I've been knocking on it for
some time with no other troubles.  Prefetch seems to me to be a little
less than ready for prime-time.  I seem to recall on a 32-bit machine
as well that pkg_delete would run terribly slowly with prefetch on,
and things got better when I re-enabled the zil and turned prefetch
off.  Not sure whether the zil just helped a lot there or if the
problem was prefetch.

I'm using ZFS on my Mac OS X laptop as well, that's been pretty good
too.  Latest build (111 on zfs.macosforge.org) has fixed mmap
coherency, and a bunch of other things.

On Fri, Apr 18, 2008 at 1:03 PM, Claus Guttesen <kometen@gmail.com> wrote:
> >  OK.  I've been able to reproduce the issue.
>  >
>  >  Conditions:
>  >  - Stock shipped kernel and modules from RELENG 7.0
>  >
>  >  loader.conf settings:
>  >  - zfs prefetch enabled
>  >  - zfs zil disabled
>  >  - vm.kmem_size_max="1073741824"
>  >  - vm.kmem_size="1073741824"
>  >
>  >  AMD64 on Core 2 Duo w/ 4 GB RAM
>  >
>  >  raidz across 4 disks, using root on zfs (also experienced this hang with zfs
>  >  just on /usr)
>  >
>  >  How To Reproduce
>  >  - install transmission-daemon, run transmission-daemon (it will daemonize,
>  >  automatically backgrounding)
>  >  - grab a torrent, such as KNOPPIX (http://torrent.unix-ag.uni-kl.de/)
>  >  - transmission-remote -a <torrentfile> (add the torrent)
>  >  - transmission-remote -s all  (start all the torrents)
>  >
>  >  wait hours to a day or so with whatever you want logging things running and
>  >  active since anything that hits disk after the hang will get hung as well
>  >
>  >  I've just done this twice.  It doesn't seem to happen with zil enabled and
>  >  prefetch off.
>  >
>  >  Expected Behavior
>  >  No hang.
>
>  Thank you for spending time on this. I'm getting some jbod-storage in
>  a few weeks and will spend some time using various block-sizes and
>  share partitions across nfs. I'm somewhat confident that zfs on
>  FreeBSD will work fine but Solaris is (of course) also an option. This
>  will be approx. 7 TB to begin with and growing upwards to some two
>  digit TB's.
>
>  What is the largest storage zfs on FreeBSD has been used on? I tried
>  it for some month on 8 TB.
>
>  --
>
>
> regards
>  Claus
>
>  When lenity and cruelty play for a kingdom,
>  the gentlest gamester is the soonest winner.
>
>  Shakespeare
>



-- 
James Snyder
Biomedical Engineering
Northwestern University
jbsnyder@gmail.com



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?33644d3c0804181115v38611102yc48c2e46f30d0bf6>