2 dhcpd Instances on Same Platform? Take 2

[Glenn Satchell]

>To: dhcp-users at isc.org
>Subject: 2 dhcpd Instances on Same Platform?  Take 2
>Date: Wed, 21 Feb 2007 14:16:01 -0600
>From: Martin McCormick <martin at dc.cis.okstate.edu>
>
>Subject: 2 dhcpd Instances on Same Platform?
>
>
>	Last Fall, our organization ran DHCP failover for a while
>and had to go back to a less robust non-failover mode because
>some wireless net authentication devices we presently have can be
>set to use only one DHCP server so traffic between the second
>server and any client got lost.
>
>	The only other problem we had was the MCLT factor since
>our DHCP environment consists of a little over 8,000 static bootp
>clients and almost as many, if not more, dynamic clients.  On any
>given day, we have to add or delete several static bootp
>addresses and the HA pair doesn't usually settle back down until
>twice the MCLT value unless one gets very lucky in which case one
>of the two would come right back up.  This would get things back
>to full normality in 1 MCLT time period.
>
>	I have been asked to research commercial solutions for
>dhcp, some of which claim not to have this problem.  I do wonder,
>however, if one could be able to run 2 instances of dhcpd from
>the same platform, using two separate configurations, of course.
>
>	One dhcpd would have not a single dynamic range defined
>for it and would only contain static bootp hosts.  The second
>instance of dhcpd would only have dynamic pools which would mean
>that one would only need to stop and restart it when the network
>configuration changed.
>
>	We are presently running dhcpd on a remote campus that is
>very small compared to our main campus and they go for weeks
>without changing their static data base.  The failover has worked
>flawlessly to the point that we had one dhcpd server die for some
>reason and our phones never rang once.
>
>	When we had failover running here, we also had no trouble
>with it except for the amount of time it takes for both systems
>to sync with each other after a kill/restart cycle.
>
>	We are wanting to get failover back again for our
>campus because we should be getting rid of the authentication
>devices that kept both DHCP servers from seeing requests from the
>wireless networks.
>
>	We don't have the luxury of being able to dedicate
>subnets to static bootp hosts and others to dynamic hosts.  If we
>did, I wouldn't need to be asking this question.  Most of our
>subnets have a DHCP range defined and everything else is static.
>
>	Any ideas are welcome.  I have been managing our ISC DHCP
>servers since 1993 and I am not looking forward to the GUI's and
>various other whizzbangs that adorn many commercial products.
>One can manage ISC-dhcpd from anywhere that will support command-line
>communications.  That's a valuable feature in itself.
>
>Martin McCormick WB5AGZ  Stillwater, OK 
>Systems Engineer
>OSU Information Technology Department Network Operations Group

I'm curious as to why you have MCLT issues when restarting the servers.

I have run several large sites with failover, and we always used the
procedure to stop and then start the secondary. Wait for it to sync
(usually only a few seconds), then stop and start the primary. Both
servers continue to hand out leases with the normal lease times after
restarting this way.

By normal I mean that the initial lease offered to a client is MCLT and
then subsequently it goes to the default lease time. But unless a
client tries to renew during the few seconds the server was shutdown
things operate normally. In fact if the client tries to renew when the
server is down, it will usually retry a short time later and because
the server is back up, things keep on working.

And it's good to see someone so dedicated to a project. Wow! 1993 must
havc been pre1.0 code... Way to go.

regards,
-glenn