[boulder.pm] error handling

[Rob Nagler]

> But times have changed, a lot, and I am wondering lazily what
> mechanisms people use for this kind of error handling, and how
> they like it.

I like calling die when a problem occurs.  I also like having a
wrapper which at some contextual information to the internal logs.

We have a few problems which result in an email being sent to root.
For the most part, we rely on high frequency log parsing.  We have a
multitiered alert system, which is controlled by regular expressions
and counters in the log parser.  For example, if we get three die
calls of a specific type, we send a page.  Otherwise, we send email.

Rob