Jonathan Steffan wrote:
There is my incomplete 2am reply ;-)
I might try to reply (or even read your post), but I'm at my 4:44am and
1) theoretical #5 doesn't work, as sda1 can't be both part of /dev/md7 AND part
of a devicemapper at the same time (even if both parts are working with
different slices). (maybe with a loop partitioning kernel patch/feature?)
2) I think I just figured out the #*(@ undocumented dm mirror target, which if
it works, was probably the obvious way to go in the first place...
If I don't happen to actually try real usage of it tonight, and someone reads
this and wants to comment on my theory of it... that would be nice...
my theory for dm mirror migration: (I did do a trivial test which seemed to work)
1) create a linear table with the original device
2) do a reload with a mirror table with the orig device first and the new device
second, then resume (this will??? work on a rootfs device?)
3) when sync is complete (oh yeah, use that flag in 2), do a reload with a
linear table of the new device.
Yeah that sure does sound simple, but could I have ever used a simple
Documentation/device-mapper/mirror.txt file that described that example. (yes,
if this works, I'll write the file and submit it)
I believe that substituting this dm mirror migration for mdadm should work, and
of course is just much better period end of story.
Though
http://sources.redhat.com/lvm2/wiki/RoadMap doesn't give me confidence
seeing the entry-
"pvmove, snapshot, mirror of root filesystem"
-dmc