https://bugzilla.novell.com/show_bug.cgi?id=682766
https://bugzilla.novell.com/show_bug.cgi?id=682766#c8
Rob Lucke changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |novell@roblucke.com
--- Comment #8 from Rob Lucke 2011-04-20 01:59:34 UTC ---
The installation partitioning tool is very confused by the Intel 82801 "raid"
device configuration. I had a single RAID1 LUN, "Volume0" comprising two
physical disks. First, the GRUB installation targeted the extended partition
containing Linux (see below), but not the MBR. Once I booted rescue and
installed GRUB in the correct place (I did not override the "automatic"
selection), I was able to get the GRUB menu, followed by a hang.
So, I did a clean install onto the RAID1 LUN comprising two 500 GB SATA disks
that already had the first primary partition occupied by Windows 7 (/boyes, I
know -- it's for work). I created an extended partition (p2) and within that
/boot (p5), swap (p6), and root (p7). The devices were visible as /dev/sda and
/dev/sdb and as the md raid devices /dev/md126 and /dev/md127.
It took several goes and some auxiliary "fdisk-ing" to get a clean partition
configuration.
The next install succeeded, but the next reboot failed. I managed to get part
way by selecting the "safe" boot (to bypass a graphics issue), and hitting ESC
at the appropriate point to see the messages. This waited for the
<diskUUID>-part" device to appear, timed out, then prompted me for fail-back to
/dev/md126p7, which succeeded in a boot, but not in mounting /dev/md126p5 onto
/boot. Without the ESC, this was not visible, and the system appeared to hang.
With a little more rooting and GRUB-ing about, I managed to find a symptom of
what is happening.
The /boot/grub/menu.lst file uses /dev/disk/by-id/<diskUUID>-<partN> as the
descriptor for the disk sections for the root and boot devices. So I should
have /dev/disk/by-id/<diskUUID>-part7 visible at boot, along with *-part5 and
*-part6. These are _not_ present in the /dev/disk/by-id directory, while the
md126 and md127 devices _are_ (correctly) present as links to ../../<device>.
The proper /dev/md126p[1-7] devices are also (thankfully) present.
It looks to me as if a udev rule (or whatever is responsible) is not properly
creating the links for the md raid device _partitions_ when they are discovered
(md126 is the active device with /dev/sda as primary and /dev/sdb as secondary)
in looking at /proc/mdstat.
By changing the stanzas in /boot/grub/menu.lst to use /dev/md126p7 as root and
/dev/md126p5 as /boot, I was able to get a mostly live system. This was after
considerable thwacking with the rescue system and mounting/changing /dev/sda7
and /dev/sda5, followed by replicating the changes to /dev/sdb7 and /dev/sdb5.
I haven't taken apart the install system pieces quite yet, but this should help
locate the problem. I hope, anyway.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
