[Pkg-puppet-devel] d2d679e004fbd73cb7f948828ea7fa947a568dcd

[Sam Quigley]

Hey --

On Apr 14, 2008, at 8:32 AM, Micah Anderson wrote:
>
> 4. Finally, I made changes to the LSB headers in the initscripts. I
> removed a long series of white space after all the headers that  
> probably
> got put in there from a cut and paste, and was causing update-rc.d to
> complain that there wasn't actually a LSB header. Secondly I got an
> email from Sam Quigley who pointed out a problem on Ubuntu gutsy which
> starts puppet before the network has been fully configured so you get
> "no route to host" errors on boot. So I added $network and $named to  
> the
> Required-Start and Required-Stop headers for the puppet initscript;  
> and
> I put $network in the Should-Start/Should-Stop headers for the
> puppetmaster. My reasoning was that the network should be there, but  
> if
> its not, puppetmaster will still run fine.
>
> However, I am wondering if maybe the puppet init script should instead
> have the Should-Start: $network and Should-Stop: $network, and not  
> make
> it a hard requirement (as the Required-Start does), but an advisable
> one. I was thinking of the case where I might be with my laptop on the
> train, with no internet, but I have a local dev puppetmaster running  
> on
> my laptop and a puppetd that queries my puppetmaster on
> 127.0.0.1... that seems like a legitimate setup and shouldn't be
> forbidden (which would be the case if using the Required fields).

I guess the question is whether Requiring $network simply ensures  
"ifup -a" has been called (which would presumably work, even if it  
didn't do anything, on a disconnected machine) -- or if it would  
actually crap out unless the network comes up.  I'd have assumed the  
former, but unfortunately, I don't know enough about the header  
semantics to answer the question with any real authority.

-sq