Broken ZFS boot on upgrade

[Jonathan Anderson]

On 11/10, Kevin P. Neal wrote:
> Do you have snapshots? If you have a snapshot from before you upgraded
> then you can roll back the upgrade. That's another way to get you back
> to a bootable system.

I do have snapshots, though mightn't the effectiveness of a rollback depend on
the reason for the failure? For example, if the issue is the 2+ TB thing,
mightn't new blocks created by the rollback also end up on the large,
not-visible-to-BIOS disks?


> If this is the case then can you remove disks one at a time and replace
> them with GPT-partitioned disks? If you are using mirrors then this may
> work if the smallest partition/disk currently in use per-vdev is no larger
> than a disk that has been partitioned.

Hmmmm... I think that, unfortunately, the mirrors are mostly exactly the same
size (four 1 TiB and two 3 TiB), so there isn't any slack for resizing.


> Sensible? Mmmmmm... I would say that multiple ZFS partitions on a single
> disk is something to be avoided if possible. There may be cases where it
> can't be avoided, but it's not the best. Multiple partitions on a disk in
> the _same_ _pool_ sounds just plain bad and I wouldn't do it.

Ok, thanks!


Jon
-- 
Jonathan Anderson

jonathan at FreeBSD.org