Anyway to change pool to use the gpt label instead of gptid?

[Larry Rosenman]

On 10/23/2011 9:16 PM, Jeremy Chadwick wrote:
> On Sun, Oct 23, 2011 at 07:01:31PM -0700, Artem Belevich wrote:
>> On Sun, Oct 23, 2011 at 6:49 PM, Larry Rosenman<ler at lerctr.org>  wrote:
>>> On Sun, 23 Oct 2011, Artem Belevich wrote:
>>>
>>>> On Sun, Oct 23, 2011 at 5:52 PM, Larry Rosenman<ler at lerctr.org>  wrote:
>>>>> Is there any way to convert this setup to use the GPT labels instead of
>>>>> the
>>>>> the GPT UUID's in zpool status/zpool iostat?
>>>> You can try exporting the pool and then importing it back with "zfs
>>>> import -d /dev/gpt"
>>> Really hard to do with root on this pool :)
>> You may be able to solve that part with LiveCD. I'm still not quite
>> sure where device info is kept -- in the pool or in
>> /boot/zfs/zpool.cache. ...
> Both.
>
> icarus# zpool status data
>    pool: data
>   state: ONLINE
>   scan: resilvered 279G in 1h0m with 0 errors on Thu Aug 11 22:43:38 2011
> config:
>
>          NAME        STATE     READ WRITE CKSUM
>          data        ONLINE       0     0     0
>            mirror-0  ONLINE       0     0     0
>              ada1    ONLINE       0     0     0
>              ada3    ONLINE       0     0     0
>
> errors: No known data errors
>
> icarus# strings /boot/zfs/zpool.cache | egrep 'ada[13]'
>          /dev/ada1
>          /dev/ada1
>          /dev/ada3
>          /dev/ada3
>
> icarus# dd if=/dev/ada1 of=/dev/stdout count=128 | strings | grep 'ada1'
> 128+0 records in
> 128+0 records out
> 65536 bytes transferred in 0.013422 secs (4882725 bytes/sec)
>          /dev/ada1
>          /dev/ada1
>
> icarus# dd if=/dev/ada3 of=/dev/stdout count=128 | strings | grep 'ada3'
> 128+0 records in
> 128+0 records out
> 65536 bytes transferred in 0.023449 secs (2794838 bytes/sec)
>          /dev/ada3
>          /dev/ada3
>
> Larry, have you considered avoiding ZFS on root and instead just using
> gmirror on your OS partition (not the entire disk, and not the swap
> partition)?  I've read that gmirror "just works" and is reliable,
> including booting from it, and supports hot spares.
>
> Thought I remember there being some caveat to swap being stored under
> gmirror -- the man page mentions caveats that relate to getting kernel
> panics/dumps -- so that may be a problem.  Using multiple partitions and
> mirroring only the UFS/UFS2 partition sounds like a better choice.
>
> I know it would be ideal to have it all controlled with ZFS, I
> understand/hear you (especially since I just got done harping on added
> complexity, re: GPT labels vs. device names), but welcome to the mess.
I just got out of a ufs+zfs setup.  And I like the idea of being able to 
use all of the disk space as ZFS, and replacing ANY
drive that dies and not losing data (yes, I know, raidz1 only handles 
one failure at a time, but that's fine).
(current is 5x400g, 1x500g disk).  I intend at some point of replacing 
them all with 2T or 3T disks, and then leave it
alone for a while.


I think I'll just deal with the adaxpy names (via 
kern.geom.label.gptid.enable="0" in /boot/loader.conf), and
just deal with it.

I think it would be nice to have the zfs/gpt/loader/boot have a knob to 
prefer the label vs the physical device name.

It is VERY unlikely that the physical name will change ON THIS SERVER, 
since it only has 6 SATA ports on it (supermicro 7045B-TR+ system  
http://www.supermicro.com/products/system/4U/7045/SYS-7045B-TR_.cfm).

Appreciate everyone's comments.  I didn't know about the tunable, which 
helps somewhat.