Re: Etch d-i beta2 : bug for lvm on raid ?
On Fri, June 2, 2006 12:10, Christophe Chisogne said:
>...
> Configuring md seems to work, but syncing is slow: /dev/md2 is about
> 200Go.
> (is there a way to avoid the resync, as it was already done previously
> btw?)
If you assemble the array manually using mdadm you probably can (see the
man page for mdadm, I think it's something like --assume-clean).
> The real problem is the lvm part. If I set /dev/md2 as physical volume for
> lvm, problems begin. d-i really wants to update the partitions on volume
> "/dev/md2", as if it was a disk, before configuring lvm. If I accept, I
> receive an error like this "errors informing kernel about modifs to
> partition /dev/md/2p1, invalid arg". This looks like a bug to me : md2 is
> not a disk, and it has no partitions. What's wrong ?
This would be a parted bug. It likes to treat all block devices it finds
as harddiscs with partitions. A patch similar to the lvm2 patch
(http://bugs.debian.org/cgi-bin/bugreport.cgi/parted-367965-lvm2-v2.dpatch?bug=367965;msg=8;att=1)
is probably needed for md.
As a long term solution, support for
block-device-with-no-partitions-present-or-possible should probably be
added to parted (for dm, md, lvm, evms, etc).
>From that point, "cat /proc/mdstat" seems to freeze. d-i tries to
>ext3-format the /boot partition, but seems to freeze.
parted probably mucked around with the underlying partitions...
> Is it a known bug ? I'm using beta2, but tested with 2005-05-28 daily
> build,
> without success (to avoid the "tr" bug of 17 May listed on
> DebianInstaller/Today)
>
> Are there work-arounds using CLI ? Perhaps I can create lvms by hand,
> and later convince d-i to use them as /usr, /var etc ?
Try to first create the RAID1 using the regular user interface, then goto
the second console and create the lv's by hand (pvcreate, vgcreate,
lvcreate)...then you should be able to go back to the d-i partitioning
menu and refresh it by exiting and re-entering it. Hopefully your lvm
partitions will show up after that.
Regards,
David
Reply to: