As I recall from before you are using a non-RPM installed version of
Cobbler so that you could put /var/lib/cobbler on some other partion?
What are the details regarding that again?
That's correct. Basically, I have run "sed" against the source code to
replace hard-coded paths to ensure that Cobbler references everything
from inside our own software tree. While I did encounter some problems
with my tinkering in the past, the software is now working as expected,
aside from this problem, that is.
The most likely cause is that cobblerd somehow crashed while updating
the state file. There are various checks in place to make sure invalid
contents don't cause problems, so I'm not sure what might be causing
this. Is /var/lib/cobbler on NFS or something, or could the system have
run out of memory?
I heeded the warning concerning NFS and locking. All of the areas where
Cobbler writes are on local disk. Distros, kickstart files, and snippets
are kept on NFS so that my both of my Cobbler servers can access them
easily. I also wanted to use file system ACLs so we can manage cobbler
without requiring the root, and ACLs do not work over NFS in our current
setup.
As for memory, the server has 8G and its processes are mostly I/O bound
(i.e. httpd, rsync, cfserv and cobblerd). As much as this sounds like a
resource problem, I don't think it is.
And yes, reporting items with the latest dot release is pretty much
requirement a reproducibility as problems of various sorts may have
already been resolved -- 2.0 is still in testing for the next week or so.
If I find some time today, I will try running the "nopxe" trigger
manually to see if I can cause it to crash. If I can figure out how to
reproduce this on cue it will at least help pinpoint where the problem
lies. Right now I need to get my primary Cobbler server back online.
Thanks.
--
Chris O'Regan <chris(a)encs.concordia.ca>
Senior Unix Systems Administrator, Academic IT Services
Faculty of Engineering and Computer Science
Concordia University, Montreal, Canada