New builds won't boot

[Chris Hedley]

Hi all,

I was wondering if anybody has any suggestions for me.  Since around 
January this year, I've been unable to get new builds to boot on my 
system: they go through the usual kernel initialisation stage, reach the 
"ZFS is an experimental feature" message, sometimes followed a few other 
random notices, and then it hangs indefinitely.

I've just regressed to a version from mid January which boots without any 
problems, though ZFS from that point is unstable enough that I'm unable to 
update any of the packages on the security alerts list without it locking 
up, so it's not ideal.

I've tried various approaches to see if I can find out what was the 
problem.  I did read that GEOM_MIRROR has displayed similar problems in 
recent versions, especially if combined with k8temp, oddly, but disabling 
one or other, or both, or making it part of the kernel, and various other 
trial-and-error configurations haven't yielded any success.  Other 
attempts have been to try a build both with and without SMP, turning off 
optimisation in CFLAGS and numerous other fiddles that I can't recall now.

A further complication is that USB appears also to be broken in the newer 
versions, so although I can break out into the debugger, the keyboard 
doesn't work from that point onwards.  Not that it makes a lot of 
difference as I'm not really sure what to look for.

It is also worth noticing that a build of GENERIC also has the above two 
problems so I'm not entirely convinced it's a configuration issue unless 
there's something weird in my loader.conf, which is as follows if it looks 
suspect to anybody:

#geom_mirror_load="YES"
geom_label_load="YES"
geom_eli_load="YES"

nvram_load="YES"

zfs_load="YES"
vm.kmem_size="1536M"
vm.kmem_size_max="1536M"
vfs.zfs.arc_min="512M"
vfs.zfs.arc_max="1024M"
vfs.zfs.prefetch_disable=1
# vfs.zfs.zil_disable=1

vfs.root.mountfrom="zfs:tank/root.2009.01"

kern.maxfiles="25000"

My system setup is a uniprocessor Opteron 248 in a Tyan Thunder dual 
processor motherboard with 3GB registered/ECC RAM.  I use ZFS for my main 
storage array in a RAIDZ2 configuration spread across eight of its 10 
partitioned SATA discs; the system boots off one component of a 
GEOM_MIRROR enabled partition to avoid any need for zfsboot and related 
issues.

I'm wondering if it might be ZFS that's the problem here.  I'm not 
especially enthusiastic about giving it up, so might I be better off 
migrating to Solaris, do you think?

Any suggestions would be welcome!