OpenZFS ZFS-OSX (osx.zfs-20130929.dmg): testing

[Graham Perrin]

Test environment: 

* OS X 10.9, pre-release

* MacBookPro5,2

* 8 GB memory

* frontmost USB 2.0 port to a hub with a two-disk pool (tall)

* central USB 2.0 port to an old keyboard

* rearmost USB 2.0 port to a hub with one disk to start OS X, the other disk is a one-disk pool (zhandy)

* another pool (gjp22) on a part of the internal sshd that's encrypted with Core Storage

* pool version 28 – the disks are normally used with ZEVO on Mountain Lion.

Tests included: 

* -o readonly=on at time of import

* forced imports (this round of tests was preceded by an intended kernel panic whilst using the pools with ZEVO)

* read-only attachment of sparse bundle disk images where the .sparsebundle is on ZFS

* exports

* normal imports

* normal attachment of disk images

* restoration of an .app (Firefox) 

– from a Time Machine backup that's stored on a .sparsebundle on a child of tall

– to HFS Plus

* launch and quit of the app

* exports. 

Test results: 

* Nothing adverse

* less than a handful of minor issues, which I can't report at this time.

Now back with ZEVO on Mountain Lion, my plans include: 

* clone the 12F37 startup volume, which excludes my encrypted ZFS home directory, to the disk on the hub that was used (above) for preliminary testing of today's ZFS-OSX

– progressing

* backup

* upgrade the clone to a pre-release build of OS X 10.9

* backup

* start Mavericks

* cautiously test my encrypted ZFS home directory with ZFS-OSX

– initial focus on usability of a jollyjinx script that I habitually use for snapshots, send, receive and destruction with ZEVO. 

[Graham Perrin]

In my eagerness to progress, I didn't bother to set aside the saved Terminal output from yesterday's test session, which began around 17:00 yesterday and probably ended before 18:00. The disk with that output was subsequently repartitioned (to be used as a clone of something). 

Later with zpool status in ZEVO: for the three pools that I had tested with ZFS-OSX, I found errors affecting one of the three. 

I'm scrubbing a different pool before shifting attention back to the pool that was affected by errors. Hoping for the current scrub to finish before my work day begins. 

-----

A key point from yesterday's use of ZFS-OSX:

> * forced imports (this round of tests was preceded by an intended kernel panic whilst using the pools with ZEVO)

With ZEVO there's automatic pool rollback (for drives that don’t flush correctly) <http://zevo.getgreenbytes.com/wiki/pmwiki.php?n=Site.Overview> but I don't know whether that would apply to the hardware that I normally use. (Is it something that just works that I take for granted? I wonder.)

I don't know whether, with ZFS-OSX, I should have done something more than simply apply force when importing the pools. Related: 

OpenZFS ZFS-OSX support for USB

<https://groups.google.com/d/msg/maczfs-devel/1MmlG7qKhgg/IMgdVyIILOEJ>

[Graham Perrin]

<http://zevo.getgreenbytes.com/forum/viewtopic.php?p=4532#p4532> reminds me that automatic pool rollback is a cross-platform feature. 

Concerning the permanent errors: I'll post in the ZEVO troubleshooting area, with reference to this topic. 

Here in MacZFS-devel there remains the question about USB.