gmirror on 7B4

Chris H. chris# at 1command.com
Fri Jan 4 01:56:28 PST 2008


Quoting Clifton Royston <cliftonr at lava.net>:

> On Wed, Jan 02, 2008 at 08:47:43AM -0800, Chris H. wrote:
>> Quoting John Nielsen <lists at jnielsen.net>:
>>
>> >I'm not sure I remember everything from earlier in this thread so I
>> >don't know if it's relevant, BUT you can't boot from a gstripe volume
>> >(or from a gconcat one AFAIK). Inferring from your fstab example
>> >below it doesn't sound like you intend to but I just wanted to be
>> >sure.
>>
>> Are you sure? I read that using gmirror requires /kernel to be located
>> in the /boot slice and everything else (all other slices) can be mirrored
>> safely. But in all my reading (man pages, FBSD handbook, asstd articles)
>> I haven't seen anything indicating booting wasn't possible from a gstripe
>> volume.
>
>  Your current idea is backwards; you can boot from entirely mirrored
> drives (i.e. RAID1) and I've been doing it since 5.3, but AFAIK it is
> impossible to boot from a striped drive and I suspect will remain so
> for a long time.
>
>  One way to visualize this is to recognize that because the gmirror
> information is stored at the very end of the lower-level GEOM object,
> each of the raw drives in the mirrored set appears to be an perfectly
> normal drive when reading it from its beginning; thus it is possible to
> simply read it as a normal device during the earlier stages of boot
> until GEOM and gmirror loads.  With striping, however, the logical
> content is spread out across multiple drives, so any one drive you try
> to boot from has only 1/Nth of the relevant sectors.

Indeed, and thank you for pointing out the obvious to me. :)
I was almost immediately reminded of that after posting. :P

But really, I appreciate your taking the time to /enlighten/ me.
It /does/ help.
Given the /wealth/ of information afforded to me here on the list,
after proposing my intentions. It quickly occurred to me that I
had developed quite a few misconceptions about GEOM and friends,
and that I should have taken just a bit more time before leaping.
In the final analysis, I think it would be /far/ more efficient if
I simply blanked my current disk, and simply laid it out as I ultimately
want it. Then simply unarc the root folders to their desired destinations
from the most recent backups. Which kind of makes this thread a loop.
As my initial question was why wasn't gMIRROR part of sysinstall.
It's funny, I've spent over 2 decades running *BSD, and yet I never
really spent much time obtaining intimate knowledge about the disk
"construction". Oh, it's not that I know nothing about it. But rather,
that once I determined the ultimate layout for my needs, I simply
let sysinstall handle it. So other than needing to add disks and move/
re-create slices, I was done. But as I now revisit it, I discover I
should probably spend a little more time acquainting myself with it. :)

Thanks again for taking the time to respond. I appreciate it.

Chris

>
>  Does this help?
>
>  -- Clifton
>
>
> --
>    Clifton Royston  --  cliftonr at iandicomputing.com / cliftonr at lava.net
>       President  - I and I Computing * http://www.iandicomputing.com/
> Custom programming, network design, systems and network consulting services
> _______________________________________________
> freebsd-stable at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe at freebsd.org"
>



-- 
panic: kernel trap (ignored)





More information about the freebsd-stable mailing list