ZFSKnownProblems - needs revision?

[Ben Kelly]

On Apr 8, 2009, at 12:18 PM, Christian Walther wrote:
> I used geli encrypted ZFS including Root on my IBM Thinkpad T31 with
> 1GB RAM on a 160GB HDD. (i386 7-STABLE)
> Swap on a dedicated slice. Some Z Filesystems used compression
> (/usr/ports, /usr/src, for example).
>
> I encountered several crashes, especially during heavy loads, such as
> compiling big ports.

Can you clarify what you mean by "crashes"?

I have a very similar system here using zfs on geli with compression  
on a non-SMP i386 machine.  The only real difference is that I am  
running 8-CURRENT.  With this setup I've found that I can reliably get  
the system to go into livelock under load.  I posted some analysis and  
a possible patch in this thread:

   http://docs.freebsd.org/cgi/getmsg.cgi?fetch=832196+0+archive/2009/freebsd-current/20090322.freebsd-current

I'm still investigating what exactly triggers the problem, but it  
appears that if the ARC cache comes under memory pressure at the wrong  
time certain zfs transaction threads can enter an interlocked looping  
pattern.  Without proper thread prioritization this can prevent the  
system from writing data out and thus keep the loop running forever.

You can find a better copy of the patch here:

   http://www.wanderview.com/svn/public/misc/zfs_livelock/zfs_thread_priority.diff

I intend to update this with some tweaked priority values, but this  
should still work fine.

Please note, however, that at least one person encountered a panic  
when they tried this patch.  Its not clear if this was due to the  
thread priority changes interacting with their zfs-on-root  
configuration or if it was unrelated 8-CURRENT changes.  Also, I have  
not tested the patch against 7-STABLE.  I'd be interested to know if  
it helps with your systems (although it sounds like you abandoned zfs  
already).

Anyway, I hope that helps.

- Ben