I am attempting an upgrade from CentOS 5.3 to CentOS 5.4.
The system I am running has three 1 terabyte drives /sda/b/c configured
as soft raid0 with xfs file system and one solid state 32 GB drive (/sdd)with
linux(CentOS 5.3) on it.
Now the problem is when I attempt an upgrade installer complains about
unreadable partition table on sda/b/c and ask for intialization, which is fine
since I can ignore the message but than it try to mount /dev/md0 soft raid
partition and fails the upgrade.
Now before attempting an upgrade if I unmount the /dev/md0 and remove the entry
from fstab, installer dosen't complain about any thing and the upgrade is
Now the company I work for ships the boxes with above configurations and I
don't want customers to playing around with fstab and than upgrade their boxes.
Any work around ....?? Apart from performing upgrade over net as it must be
One way might
be to include a small pre-upgrade script on your DVD. Ensure your customers
run the script as part of the upgrade procedure, then the script can configure
fstab and reboot.
complicated way might be to get the %pre section of kickstart to detect and mount
the partitions, and modify fstab from there. You would have to do that
manually because i) anaconda’s partition detection and mounting code is not
available at that point (any anaconda devs reading – that would be a useful
thing to have) and ii) it doesn’t work for you anyway. The thread CentOS upgrade - access to chroot in %pre? from
August has some examples of possible techniques to do that.
"To err is human. To purr, feline"