Tested a bunch of ZFS with 9.0_BETA [was Re: State of ZFS in 9.0_BETA]

[Brad Spencer <brad%anduin.eldar.org@localhost>]

My $DAYJOB puts me in contact with ZFS on SmartOS (Joyent private cloud)
and a T4 Sparc with ZFS, so I was quite pleased to hear that ZFS on
NetBSD had made progress.

Nearly everything I tried with ZFS on NetBSD 9.0_BETA worked as
expected.  I did the following:

1) Made and destroyed raidz and mirror pools
2) Made and destroyed snapshots
3) Made and (tried) to destroy clones
4) Did send and receives between NetBSD and NetBSD and between NetBSD
and FreeBSD in both directions.
5) Tried to use hot spares
6) Abused resilvering (this was all on DOMUs, and I simply xl
block-detached the devices)
7) Tested compression and deduplication a bit

What did not work with the above tests:

a) Hot spares...  I could not get a hot spare to actually be used.  I
could define it in the pool, but no amount of zfs replace would cause it
to be used as a replacement and nothing automatic happened.  Further,
while the man page says that autoreplace should be a attribute but it
does not appear that attribute is actually allowed.  I could unjoin the
spare from the pool and use it then use it with zfs replace, but nothing
else worked.

b) Trying to promote a clone would result in the "zfs promote ..."
command getting stuck in tstile, which mostly results in the inability
to get rid of a clone snapshot once you have one (and of course to
promote a clone to a full filesystem).  When the DOMU was forcefully
stopped and rebooted, the zpool appeared to be ok and passed scrub.



A lot of good work was obvious done to get ZFS to this state.  A big
thanks to all who did that work.



-- 
Brad Spencer - brad%anduin.eldar.org@localhost - KC8VKS - http://anduin.eldar.org