I tried to do an upgrade to 6.1, and while I have a fairly
straightforward RedHat 6.0 installation, the partition setup
is a little funky -- my root partition is /dev/hda7, and
when it gets to the part where it looks for the previous
RedHat installation it gives an error message about mounting
/dev/hdd1 (which is actually half of a RAID-0 setup), and
then bombs out with some python exceptions.
If there was a way I could *tell* it where the old
installation was, all would be cool, but I don't see any way
to do that...
See also bug #5511. These are definitely related.
I have a workaround for this now, but it's a bit ugly because there
are in fact TWO problems! First, when looking for the previous
install, anaconda tries to mount anything that is labeled as a ext2
partition --- and if the mount fails the exception stops the install
cold. That seems like a huge overreaction to me -- if the mount
fails, you should just ignore the partition. With my RAID-0 striping,
it failed the mount, but that's not a big deal -- that's obviously not
my root partition, so it should just ignore it and go on.
The work-around: relabel (temporarily) all such partitions as
something obscure (I used "amoeba") so they are ignored.
Unfortunately, that didn't fix my problem because there is a second
problem caused by a bug in the anaconda code. In particular, there's
a case statement where a "break" is missing, so NTFS partitions are
mistakenly identified as ext2 partitions! So the mount once again
fails, and the exception stops the upgrade. I'll send a patch that
fixes this, but for now the workaround is the same as before:
temporarily relabel the NTFS partition(s) to something other than NTFS
or ext2, and then try again...
See also <a href="http://developer.redhat.com/bugzilla/show_bug.cgi?
id=5555&BUGLIST=5555">bug #5555</a> for more info on this. It
appears to affect all systems with existing NTFS *AND* HPFS (OS/2)
partitions. This is probably because (I think) NTFS and HPFS utilize
the same partition ID type.
Discarding as bug contains information also found in #5555 and #5511;
see those bugs for more information