I've been able to successfully kickstart CentOS 3.9 from the base repo, but no such luck after merging os and updates. I think I did everything right - updated base/comps.xml and regenerated hdlist/hdlist2. Updated the yum repo as well although I'm pretty sure it's not used by anaconda.
This is difficult to debug. The setup is running under kvm on a CentOS 6.4 host, and for some reason, a completely headless install fails (--graphics none; virt-viewer complains loudly and dies). A graphical install with vnc looks quite different from newer OSes' ks installs, there is no VC showing package install progress or an interactive shell. I.e. I cannot tell what exactly fails.
Yes, there is a specific legacy reason for using 3.9 here.
Some updates.
Rebooting after kickstart, the vnc screen shows
Booting from Hard Disk... GRUB
It's hung there. This makes me think that grub wasn't installed correctly. But this looks normal:
# file -s /dev/dm-6 /dev/dm-6: x86 boot sector; GRand Unified Bootloader, stage1 version 0x3, boot drive 0x80, 1st sector stage2 0x882ee7, GRUB version n.m; partition 1: ID=0x83, active, starthead 1, startsector 63, 20964762 sectors, code offset 0x48 #
Next, with some kpartx magic, I'm able to mount the single partition on the lv that provides the vm's disk device and find that /boot is all kernel pieces. Modules are missing, too, so the kernel didn't get installed. Manually extracting the kernel rpm and putting the pieces into place does not help, even if I build an initrd.
-
isdtor