still LVM2 failure in linux-2.6.3-2.1.253

Lars Kellogg-Stedman lars at oddbit.com
Thu Mar 18 16:26:42 UTC 2004


> I just tried the latest  linux-2.6.3-2.1.253.  Updated everything.
> 
> I have lvm2/ext3 root.
> 
> I get:
> 
> no such device or address while trying to open /dev/Volume01/LogVol00 (this
> is my lvm root).
> (drops into root shell for repair)

I've just installed 1.90 on a system with /boot on a RAID1 partition and
everything else on LVM on top of RAID5.

I'm seeing similar problems:

(1) vgscan appears to hang.  Bugzilla #117318 implies weird interactions
with raid1 devices, but my raid1 device is *not* part of the volume group.

(2) The problem with vgscan appears to be related to a problem with pvscan.
Haven't really had the chance to investigate this more thoroughly.

(3) filesystems falsely appear to be mounted.  That is, 'df' will show
mounted filesystems, when in fact they failed to mount.  This causes
(obviously) all sorts of interesting problems.

(4) The /dev/mapper/... device nodes appear to not be created correctly
("at all").  Should we calling 'dmsetup mknodes' in rc.sysinit?  Or is this
related to (1)?

(5) Anaconda has real problems if you attempt to install onto a disk with
existing LVM volume groups.  It will crash just before the "formatting
filesystems" portion of the install because it is unable to remove an
existing volume group (I had to manually 'dd if=/dev/zero ...' all of my
partitions -- so that the installer wouldn't auto-detect them -- to get the
install to work).

-- Lars

-- 
Lars Kellogg-Stedman <lars at oddbit.com>






More information about the fedora-test-list mailing list