Status update - ZFS "stalls" and other misbehavior

[Karl Denninger]

Those of you who have seen misbehavior of various sorts while running 
-STABLE w/ZFS filesystems may want to check out the developments in this 
thread on Bugzilla.

Work has continued on the candidate patch I posted back in March, and I 
recently put forward an updated version incorporating changes made by 
Steve as well that, at least in my testing and (thus far) on one 
production system here, have completely eliminated the problem.

More eyes looking at it and running it down are obvious better so if you 
have an interest in this issue or have had problems in this area 
(/*especially*/ if you have had system "stalls") please check it out.

Those who are using my previous patch from back in March definitely want 
to look at this; while the March changes did largely fix the issue they 
did so by moving the "danger zone" further out but did not prevent the 
ZFS code from invading system RAM to the point that it could produce 
stalls -- it just made it much less likely.  The most-recent patch, as 
far as I've been able to determine, completely prevent it even when the 
system is severely overstressed in terms of demanded I/O rate or when 
the target pool is compromised (e.g. during a resilver operation.)

Thanks.

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=187594

-- 
Karl Denninger
karl at denninger.net <mailto:karl at denninger.net>
/The Market Ticker/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2711 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20140914/004f27fc/attachment.bin>