Re: another mysterious segfault

[Ed Brown]

On Mon, 2005-08-29 at 17:20, Lars Damerow wrote:

> Thanks for the tip! I'll try it out to see if it'll help with this particular
> problem, but sadly it won't help me in the long term; the number of machines 
> we
> have precludes any interactive steps in our machine buildouts.

Of course we hope that all our kickstarts run through, that's the reason
for kickstart, a scripted install capability.   But error-checking and
subshell breakouts don't MAKE the script interactive, unless there is a
problem.    You asked for a better way to debug a problem during
kickstart (specifically, a problem with cfengine).   Somewhere down the
road when there's a DNS glitch, or a network connection is down, or a
needed file on the http server (update.conf?) gets its permissions
changed, or cfengine segfaults, or one of the seemingly infinite other
reasons things don't work as planned, it's SO much easier to go to the
console on that failed node, figure out what went wrong, fix it or work
around it right there and finish the install than it is to debug
indirectly, starting the install over and over again, hoping you fixed
the problem. 

Error-checking is just good programming of course.  And writing a
kickstart script in this way makes it easier and quicker to arrive at,
and maintain, a working kickstart script.  

-Ed