Vinum migration 4.x-5.4

[Stijn Hoop]

On Fri, Aug 19, 2005 at 11:01:55AM -0500, Robin Smith wrote:
> There seems to be a consensus in the references I've found that vinum
> is completely broken on 5.4

That is true. IMHO it should be removed from RELENG_5 and _6 if it isn't
already.

> and that gvinum/geom_vinum is not ready for production use.

Well the only reason it might not be is that it hasn't seen widespread
testing, as far as I can tell it should all just work. I do use gvinum
on a 5-STABLE host and it has worked well for me in the past [1].

> As it seems to me, this means that anyone using
> 4.11 (say) and vinum will have to abandon vinum (i.e. quit doing software
> RAID) in order to upgrade to 5.4.

5.4 does have alternatives to vinum (which is another reason why gvinum
hasn't received as much testing): gmirror, graid3, gstripe, gconcat.

> That can be both laborious and slow
> (e.g. if you have /usr on, say, a four-drive vinum volume in 4.11, you're
> going to have to replace those drives with something else in order to go
> to 5.4.

I'd say building a new test box is about the only sane way to do it.

> Is that false, and is there a relatively simple way to get 
> geom_vinum in 5.4 to read a vinum configuration produced under 4.11 and
> start the vinum volume as it is?

As far as I can tell, it should just work. To pick up the latest round
of vinum fixes it might be best to run 5-STABLE (ie. RELENG_5) but it
should not be necessary unless you run into difficulties.

But the only way to know for sure if things work, is to test...

--Stijn

[1] for some reason I discovered a configuration problem earlier this
    week, but the other part of the mirror is holding up and it seems
    that I can reconstruct the broken part this weekend. If anything,
    it seems that a gvinum mirrored plex is robust.

-- 
"Coca-Cola is solely responsible for ensuring that people - too stupid to know
not to tip half-ton machines on themselves - are safe. Forget parenting - the
blame is entirely on the corporation for designing machines that look so
innocent and yet are so deadly."
	-- http://www.kuro5hin.org/?op=displaystory;sid=2001/10/28/212418/42
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20050819/371435ca/attachment.bin