I guess I'll keep following up my own posts until someone else chimes in :-). Seriously, though, had anybody else seen this problem? Has anyone successfully mounted an NFS volume from kickstart wiht CentOS 5? To recap the problem, I have the following commands in the kickstart %post section:
%post mount centosmirror:/local/www/html/ks /mnt /mnt/ks/post/post.sh > /root/post_install.log 2>&1 umount /mnt
It's pretty straight forward, and this has worked for me for years. But not anymore with CentOS 5. The mount command fails with a "mount.nfs: Input/output error" message. Here is an annotated copy of a console session after waiting for the kickstart installation to complete, and then getting to the console using "Ctrl-Alt-F2":
[1] # /usr/bin/mount centosmirror:/local/www/html /tmp # umount /tmp # ls -l /usr/bin/mount lrwxrwxrwx 1 root 0 9 Apr 10 15:41 /usr/bin/mount -> ./busybox # cp /usr/bin/busybox /mnt/sysimage/tmp [2] # /mnt/sysimage/bin/mount centosmirror:/local/www/html /tmp mount: wrong fs type, bad option, bad superblock on centosmirror:/local/www/html, missing codepage or other error In some cases useful info is found in syslog - try dmesg | tail or so
# chroot /mnt/sysimage/ [3] # /tmp/busybox mount centosmirror:/local/www/html /mnt # umount /mnt [4] # /bin/mount centosmirror:/local/www/html /mnt mount.nfs: Input/output error
[After a reboot of the newly installed CentOS 5 system]
[5] # /bin/mount centosmirror:/local/www/html /mnt
The mounts marked [1], [3], and [5] worked, but [2] and [4] didn't, even though they use the exact same binary as [5]. Note that I had to use /tmp as the mount point in steps [1] and [2], because I would have clobbered my environment if I hadn't (I learned this the hard way :-).
Conclusion: the native CentOS 5 mount command fails in a kickstart environment (with two different failure modes depending if you are chrooted or not).
I will probably work around this issue by using wget to copy all the files I need, but I still would like to figure out the root cause. Is this worth filing a bug for (either CentOS or upstream or both)?
Alfred