How do we clear a bogus zpool?

Warren Block wblock at wonkity.com
Tue Aug 20 23:40:21 UTC 2013


On Tue, 20 Aug 2013, Sean Bruno wrote:

> On Tue, 2013-08-20 at 12:41 -0400, Sam Fourman Jr. wrote:
>>>         zroot                     UNAVAIL  insufficient replicas
>>>           raidz1-0                UNAVAIL  insufficient replicas
>>>             17925463268209287656  UNAVAIL  cannot open
>>>             11020448220822113890  UNAVAIL  corrupted data
>>>             10143858893287711942  UNAVAIL  corrupted data
>>>             7542790596970715955   UNAVAIL  corrupted data
>>>             10811885036534933813  UNAVAIL  corrupted data
>>>             13343774937261906429  UNAVAIL  corrupted data
>>>
>>> # zpool destroy -f zroot
>>> cannot open 'zroot': no such pool
>>> # zpool clear -F zroot
>>> cannot open 'zroot': no such pool
>>>
>> gpart -F destroy /dev/adX
>
> Sure.  But, what device to destroy?  Its not clear from the zpool list
> where the "pool" is coming from.

gpart probably won't be able to recognize ZFS metadata and clear it. 
Would be kind of handy, though.

The man page for 'zpool labelclear' says it takes a device for a 
parameter.  But the device must not be part of a pool...

If all else fails, using dd to erase the first and last couple of 
megabytes of each of the disks ought to clear it.  (Untested...)


More information about the freebsd-fs mailing list