Lost pool after nas4free configuration asynchrony. [x-post]

[Johan PONIN]

pre-disclaimer, I sent this onto zfs-discuss before realizing it was a
zfsonlinux ML, so I'm copying the mail here.

Hi,

Disclaimer, this is my first NAS, I barely know FreeBSD, have a
computer engineer degree though and archlinux as a daily driver.
Consider this a mostly PEBCAK story. I hope you'll read through
nonetheless.

Setup
 - NAS4Free from a 4GB usbkey on HP N54L.
 - 3x1TB WD Red as a vdev
 - 1 pool, 1 dataset, raidz (single parity AFAICR).

The usbkey failed, nas4free config was lost. I `dd` it onto a new 8GB
usbkey, boot and got to the console to reset factory config to gain
access to the webgui (I remember doing that once causing no issue with
the zfs pools and datasets)

The webgui showed no more pool/dataset, but running /etc/rc.d/zfs
start mounted everything fine. zfs scrub reported no errors. Data was
there, a few `less /mnt/../file` showing good data.

Trying to recreate the pool config in the gui to avoid mounting by
hand, I got some errors, but the GUI told me it detected a pool
config, offered to synchronize but I was afraid not knowing in what
direction it would do it. I removed the pool I created. Pool and
dataset still accessible. I reboot, try a second time the "same"
thing. But now the new pool is validated, the old pool isn't detected,
and I got an empty dataset.

I tried some commands (shouldn't have, idiocy). That said, I didn't
add data to this new pool/dataset, and so I'm hoping the actual zfs
object graph is intact. Of course I don't know how deep 'zpool ...'
commands change things, that might be very naive of me.

What options are there ? Could I copy the 3 1TB blocks onto another
drive and attempt some form of uberblock scan ?

Thanks in advance for your time.

ps: what are other knowledgeable people ? freebsd-fs ML ? I have read
George Wilson's name but it seems impolite.