Re: cfengine runs every hour by default?

[cmustard]

On Fri, Apr 09, 2004 at 01:00:01PM -0500, Brendan Strejcek wrote:
> Maybe a push model is more appropriate for your problem. It would be
> overly simplistic to say that relying on cfrun is error prone, but,
> for example, what happens when one of your machines is down during the
> friday push? 

- well we also run nagios and have to know what's up or not, and why. 

> Do your hosts also poll for changes upon boot? 

- yes, we have a script that runs in initial runlevels, which
  asks cfservd server for current /inputs/*.conf files recieves them,
  then runs them locally. 


> If the previously described machine was brought back up on saturday,
> would you have an out of date machine for a week? Etc.

- should be taken care of with above scenario, but this is a good point.
  I guess there is an assumed level of human interaction in the 'push' model.

 
> I would not worry about the network traffic. As long as your clients
> are not DDoSing the cfengine master, all is well.

- I guess my big issue with our model/usage vs cfexecd, is I don't like the
  idea of having a working prod env, then something 'odd' happening to the
  cfservd and it pushing something 'bad' out without our knowledge,
  (automatically) on to previously working clients. (I realize this is an
  'issue' anyway and is one of the main benefit's of a tool like this.
  You can configure/'mis-configure' many hosts simultaneously ;) 
  

  Thanks again for the help and advice.