ZFS: /system/zones "respawning" on rpool

Hi,

I have this fresh installation of Solaris 11.3 sparc.
I have two zfs pools both using two disks in mirroring mode, both are online.

I want to move /system/zones, currently rpool/VARSHARE/zones, from rpool to the other zfs pool so my zones don't consume space on the disks allocated to rpool. Since it's a fresh install, /system/zones is currently empty, no zones have been created yet.

I boot the system in signle user mode
Then destroy rpool/VARSHARE/zones:

zfs destroy rpool/VARSHARE/zones

create the new volume and set mount point:

zfs create -o mountpoint=/system/zones storage/zones

After rebooting the system, rpool/VARSHARE/zones is automatically recreated and I have this error message:

cannot mount 'zpool/zones' on '/system/zones': mountpoint or dataset is busy
... transitioned to maintenance

If instead of destroying rpool/VARSHARE/zones I change it's mountpoint to none, everything seems to work as wanted.

So here's the question, what am I doing wrong? why is rpool/VARSHARE/zones respawning when destroyed?

Thanks and sory for the long post.

so here's the way we do this:

we have two different zpools:
rpool which is a simple mirror of disk id 0 and disk id 1

then we have another pool called zones which is a mirror of disk id 3 and disk id 4
then we just have zfs fs underneath that for each zone.

we are using Fibre Channel luns exported the host and then building zpools from the luns.

our setup is slightly more complicated than that because we are also using LDOMs. but I don't want to confuse the message. To migrate the zfs filesystems do a zfs send and receive:

Sending and Receiving ZFS Data - Oracle Solaris ZFS Administration Guide