dhcp 3.0.5 crashes with lots of lease imbalance messages

[sthaug at nethelp.no]

> This is usually a symptom of a lease database inconsistency.  A
> short-term fix will be to 'fault' the database on one of the servers
> so the peer picks up a fresh copy and they have a consistent image to
> start with.  It may slowly drift, I can't remember what we fixed after
> 3.0.5.

Running various versions of 4.1.x in a failover pair, we've never
really had lease database inconsistencies, with two exceptions:

- One inconsistency that we introduced ourselves (and thus we have
nobody else to blame :-)

- We use "stash-agent-options true" on the servers in the failover
pair. We sometimes see the option 82 info getting out of sync on
the servers. We *think* this is due to clients being moved (from
one interface to another, possibly from one router to another) in
combination with stash-agent-options *and* the option 82 info not
being transmitted in the redundancy traffic between the servers.

Steinar Haug, Nethelp consulting, sthaug at nethelp.no