segfaults when running from kickstart file

[Lars Damerow]

Hi all,

I know this has come up before, but the answer I found in the archives (don't
do it) isn't optimal for me. ;)

I'm trying to run cfengine on our machines during the %post step of our
kickstart process. More specifically, %post calls a shell script that then
repeatedly calls cfengine.

On some types of our machines, it works like a charm. On others, though, it
consistently fails in the same way: a segmentation fault after an _llseek call.
I've attached an strace. The baffling thing is that, when I alt-F2 to a shell,
chroot /mnt/sysimage, and then run the same cfengine script by hand, it works
just fine.

Does anyone have any clues as to why this might be failing? I've verified that
the kickstarting machine's IP is in the range allowed by the server and that,
when I run the script manually, the server is accepting the machine's public
key.

I really don't want to resort to running the cfengine scripts after the machine
reboots, since the changes the scripts make require another reboot. These
reboots add up when I have to do them for 1500 machines. :P

Please CC: me on any responses, as I'm not a member of the list.

Thanks for your help!
-lars

___________________________________________________________
lars damerow
button pusher
pixar animation studios

"Nauseous. Nauseated. The first means 'sickening to contemplate'; the second
means 'sick at the stomach.' Do not, therefore, say, "I feel nauseous," unless
you are sure you have that effect on others."

- William Strunk Jr., "The Elements of Style"

cfengine.strace
Description: Text document