[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Status of LVM/RAID in d-i?



Frans Pop wrote:
> I'm just looking at the changes in LVM and RAID and I'm wondering what the 
> status is. How well has the new code been tested?

I had very serious issues with lvm management (Etch d-i beta2 and 2006-06-12).

My config looks like the following.

RAID 1 /dev/md0 (/dev/sda1, /dev/sdb1) on /boot
RAID 1 /dev/md1 (/dev/sda2, /dev/sdb2) on /
RAID 1 /dev/md2 (/dev/sda3, /dev/sdb3) as lvm volume (lvs for /usr, /var...)

By not using root on lvm on raid, I naively thought that it'll avoid
boot problems etc, but I was very wrong.

1. partman wanted to update the partition table on /dev/md2,
   which seemed to lead to big problems (like cat /proc/mdstat
   hang, then d-i hang/crash) So I guess lvm on raid is not (well?)
   supported by current d-i. (Resolved by manually creating PV, VG, LVs)
   If it's unsupported, it's my fault. But then I think partman should
   make it impossible to create lvm on raid on partitions.

2. I also had problems because /boot and / were on different partitions
   (grub and lilo both failed). I used a separate /boot to avoid those
   kind of problems ; bad idea apparently. Resolved by reinstalling
   (from scratch) with /boot and / on the same partitions.

3. On another box, an initrd bug prevented the box from loading
   (could not find VG, because of missing ide/dm-mod modules)
   (Resolved by manually editing /etc/mkinitramfs/modules
   and dpkg-reconfigure linux-image-2.6...)

>From my (nightmarish) experiences (which I really dont want to repeat),
I'm sorry to say that lvm/raid support is quite far from production
quality right now. But I'm sure this could be improved in the future betas.

Regards from an unlucky Debian user,

Christophe

PS Another unrelated boot issue: some Etch upgrade replaced "auto eth0"
   with "allow-hotplug eth0" in interfaces(5), which resulted in 
   a box without network after reboot. Not sure it was a good idea.
   Is it a bug or a feature ? (Resolved by using "auto eth0")




Reply to: