[OmniOS-discuss] ZFS on iSCSI Target. OmniOS as Initiator

[Doug Hughes]

On 2/27/2014 10:56 AM, Dan Swartzendruber wrote:
>> If you have the pool on the raw device (i.e. you haven't created any
> slices/partitions and instructed zfs to use the whole 'disk'), and the
> device suddenly appears to the OS with a larger size, then ZFS will use
> all the space available. This also applies to physical drives, where if
> you replace all the mirrored drives in a pool or top level vdev with
> larger drives, the pool will suddenly be larger.
>
> Mark, you are talking about a scenario where a pool has redundancy.  You
> replace a disk with a larger one and it resilvers, using the new size.  He
> has (it seems) a single drive in that vdev.  By definition, you can't
> replace it without losing all the data, no?

Are you taking into account this option from the zpool?
zpool1  autoexpand     off                  default


if you set autoexpand = on, you will see the behavior that replacing all 
of the disks will autoexpand the pool. This can be unfortunate if the 
disk model has been slightly tweaked (happened to us) for the same exact 
disk and shows to the OS as slightly fewer blocks. It won't resilver. I 
usually only set autoexpand = on if I explicitly want to grow the size 
of the pool, and then turn it back off again right away.