ZFS hang issue and prefetch_disable - UPDATE

Oliver Fromme olli at lurza.secnetix.de
Wed Aug 6 18:07:36 UTC 2008


CZUCZY Gergely wrote:
 > The thing, that it works on pointyhat, and it works on kris@'s box, is just an
 > IWFM-level, not the proof of any stability, reliability.

You cannot "prove" stability or reliability.  If you think
you can, please tell me how.

 > No matter how much someone tunes ZFS, no matter what you disable, it's not
 > garanteed, [...]

You want a guarantee?  There is none.  Not for ZFS, not for
UFS, not for any other file system on any operating system,
be it commercial or open-source.

I agree that ZFS still has problems, especially related to
kernel memory.  AFAIK this is being worked on, and Pawel's
latest patches seem to improve things a lot.  Note that the
ZFS code is still considered experimental (you've seen the
fat warning, I assume), so it's reasonable to expect that
it doesn't provide production-quality yet.  It is rather
totally unreasonable to expect a port of ZFS to appear in
FreeBSD and be bug-free and without problems from day one.

Also note that, in earlier days, certain UFS-features such
as soft-updates and dirhash were also known to need a lot
of memory (well, "a lot" by standards as of those days),
and it's still wise to disable them on small embedded boxes
with limited RAM.

The memory requirements of ZFS aren't that much different,
although on a larger scale.  Alan Cox has worked on the
problem and lifted the existing kmem limit for amd64 in
FreeBSD -current.  (I'm not sure if he will MFC that to
7-stable.)  If you run with that code *and* Pawels latest
ZFS patches, you should be a lot less likely to see the
dreaded kmem panics.  And that's without any tuning.

Of course it might still make sense to tune ZFS for your
workload in order to get better performance (e.g. disable
prefetch etc.).  Just the same as UFS.

 > Many of us has reported this, bot noone looked into it.

That's completely untrue.  Some people put a lot of time
and efforts in FreeBSD's ZFS port.  Please don't insult
them.

Best regards
   Oliver

-- 
Oliver Fromme, secnetix GmbH & Co. KG, Marktplatz 29, 85567 Grafing b. M.
Handelsregister: Registergericht Muenchen, HRA 74606,  Geschäftsfuehrung:
secnetix Verwaltungsgesellsch. mbH, Handelsregister: Registergericht Mün-
chen, HRB 125758,  Geschäftsführer: Maik Bachmann, Olaf Erb, Ralf Gebhart

FreeBSD-Dienstleistungen, -Produkte und mehr:  http://www.secnetix.de/bsd

"C++ is over-complicated nonsense. And Bjorn Shoestrap's book
a danger to public health. I tried reading it once, I was in
recovery for months."
        -- Cliff Sarginson


More information about the freebsd-fs mailing list