On Thu, 2004-05-13 at 23:07, Steve Langasek wrote: > On Thu, May 13, 2004 at 11:26:16AM -0400, Andres Salomon wrote: > > On Thu, 2004-05-13 at 11:16, Martin Michlmayr wrote: > > > * Andres Salomon <dilinger@voxel.net> [2004-05-13 11:11]: > > > > Yes, it works for all kernel versions using device-mapper. Does d-i > > > > allow you to pick whether to use lvm10 or lvm2? If so, then you'll need > > > > to keep your lvm10 check for /proc/lvm, and add the additional > > > > /proc/misc dm check for lvm2. If d-i always uses lvm2, then checking > > > > /proc/misc (regardless of kernel version) should always work. > > > > We always use lvm2 now... I don't think it's worthwhile to give offer > > > lvm1 as an option. Do you disagree? > > > It depends if users want snapshotting support. If not, then there's no > > need for lvm10. If users do want snapshotting support, then we're left > > with a few options: > > > - support choosing lvm10. > > > - get a more complete device-mapper patch into the debian 2.4 kernel > > (current debian 2.4 only includes a subset of the device-mapper patch). > > This is something I'd push for with whomever Herbert hands the kernel > > over to. For 2.6, snapshotting is not an option yet. > > > - get a more complete dm patch into the debian 2.4 kernel, and include > > the experimental 2.6 dm patch that includes snapshot support. Not > > recommended. > > Even if someone does want snapshotting support, I can't imagine this > being relevant within the installer. They can always upgrade to a > snapshot-capable lvm2 kernel once it becomes available. > > Cheers, I was thinking more along the lines of someone installing 2.6 w/ d-i, attempting to take a snapshot, and being suprised when it doesn't work (despite docs in the package saying that it should work). Then, in order to get snapshots working, they realize they'll need to use an experimental devmapper patch. If the system is destined for production (hey, I know people using 2.6 in production already), that probably isn't an option. They can't downgrade to lvm10, either, since lvm2 LVs will have been created with lvm2-only metadata. They must downgrade to a self-compiled 2.4 kernel, with proper device-mapper patch applied manually. If I was a user, this would probably piss me off. :p Of course, this problem gets solved quicker w/ more people testing (experimental) 2.6 devmapper snapshotting patches.
Attachment:
signature.asc
Description: This is a digitally signed message part