Bug#504397: For the record
12:00 < tbm> cjwatson: do you know what the best way is to solve #504397, i.e. how to match a md partition to the underlying physical partition
12:00 < tbm> (md could be done with /proc/mdstat I guess, but I'm looking for a more generic solution in partman)
12:02 < cjwatson> oh god. err if you want a more generic solution then we need to pull our fingers out and completely remodel partman-{lvm,md}
12:02 < cjwatson> at least IMO
12:02 < tbm> cjwatson: so for now the way to go is special case md devices and look up mdstat?
12:02 < cjwatson> complex block devices are not really represented properly in /var/lib/partman/devices/ right now so I think everything ends up going through mdstat and similar hacks
12:02 < cjwatson> yes, I think so unfortunately
12:03 < cjwatson> I'd like to fix this since it would let us operate predictively on complex block devices without having to have those intermediate commit steps
12:03 < cjwatson> the obvious stumbling block is predicting the sizes of resulting block devices before running the tools
12:03 < tbm> ok, thanks
12:04 < cjwatson> e.g. an lvm vg is going to have a smaller size than the sum of all the block devices with pvs on them due to the presence of lvm metadata
12:04 < cjwatson> and you have to get all that exactly right
12:06 < cjwatson> if there's common code being used to do this mapping in several places, it might be good to put it in /lib/partman/lib/ maybe in partman-md
--
Martin Michlmayr
http://www.cyrius.com/
Reply to: