DHCP failover not balancing

[David W. Hankins]

On Tue, Sep 02, 2008 at 12:44:53PM +0800, li jun wrote:
> and  also  the same problem troubles me for a long time 
> would anyone help me
> 
> > Primary server:
> > Aug 20 09:17:04 primary dhcpd: balancing pool 8123240 1.2.3/22  total
> > 949  free 206  backup 382  lts -88  max-own (+/-)59 Aug 20 
> > 09:17:04 primary dhcpd: balanced pool 8123240 1.2.3/22  total
> > 949  free 206  backup 382  lts -88  max-misbal 88
> > 
> > Secondary server:
> > Aug 20 09:17:04 secondary dhcpd: balanced pool 8123178 1.2.3/22  total
> > 949  free 588  backup 0  lts -294  max-misbal 88 Aug 20 
> > 09:17:04 secondary dhcpd: balancing pool 814e9f0 1.2.3/22  
> > total 30  free 19  backup 0  lts -9  max-own (+/-)2  
> > (requesting peer rebalance!)

This sort of problem is a straight database inconsistency ("backup 0"
is the key).  Additionally, I'm concerned that for 1.2.3/22, it seems
the primary and secondary are not consistently configured (the 'total'
count is way off).

To get your servers back up and running, you want to "fault" the lease
database; in this case the secondary's.

  kill dhcpd
  mv dhcpd.leases dhcpd.leases.save
  touch dhcpd.leases
  dhcpd [flags/options]

The servers should cycle through recovery and get back to operational.

I'm tracking a 3.0.x->3.1.x migration problem with lease database
consistency, we'll publish a fix in the next maintenance release, but
I'm not aware of any recurring/persistent problems here (yet?).

To fix the configuration inconsistency, the BCP on failover ops right
now is to run a dhcpd.conf on each server, and a dhcpd.include.conf
for the subnet/etc configs, then include it from the server-specific
dhcpd.conf's.  Just scp the included dhcpd.conf between the servers
when you reconfigure.  This way you are very well assured they are
both loading the same config.

-- 
Ash bugud-gul durbatuluk agh burzum-ishi krimpatul.
Why settle for the lesser evil?	 https://secure.isc.org/store/t-shirt/
-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		     you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins