[dns-operations] DNS zone monitoring

[Matthew Pounsett]

On 14-Jun-2010, at 05:34, Joe Greco wrote:

> That's sufficiently obvious that now I almost feel silly.  It's  
> still not
> quite what I'd prefer, but I think it addresses many of the cases  
> that I
> can think of.  Others, like disk-full-failed-to-transfer, haven't  
> actually
> shown themselves to be likely scenarios...  maybe that could be  
> tested by
> shifting the length of the test zone file around (yes, a crappy/ 
> hacky test,
> obviously).

A disk full error should be caught by your other monitoring  
(preferably setting off an alarm well before it's actually full).  I  
tend to avoid unnecessary duplicate monitoring; if I'm already alerted  
to disk full errors, then I don't need to be alerted to the side  
effects of a full disk since I can infer those.  Setting things up  
that way avoids the desire to launch my phone into orbit when a single  
failure generates dozens of alerts from the cascade of side effects.

Matt