Failover and config differences

Chuck Anderson cra at WPI.EDU
Wed Dec 12 12:35:16 UTC 2012


On Wed, Dec 12, 2012 at 12:28:32PM +0000, Phil Mayers wrote:
> One is that, as noted, we re-build and reload the DHCP config quite
> frequently. 20-30 times a day might not be unusual, depending on the
> number of updates into the database.
> 
> I was never clear how (if at all) failover handled the config being
> different at each end. For example, suppose I have a config
> fragment:
> 
> network x netmask y {
>   pool {
>     192.168.1.10 192.168.1.20;
>     failover peer xxx;
>   }
> }
> 
> ...and the pool changes for whatever reason:
> 
> network x netmask y {
>   pool {
>     192.168.1.10 192.168.1.18;
>     # 192.168.1.19 is now gone
>     192.168.1.20;
>     failover peer xxx;
>   }
> }
> 
> ...and one DHCP server re-starts fractionally (say a minute, for
> example) before the other. What happens?

We've been doing DHCP Failover for many years, restarting every 20
minutes as necessary just like you describe, and we've never had any
issues due to config differences for a minute or two between the
failover peers.  We've added/deleted pools infrequently and never had
a problem.

> Related: what happens if I make an OMAPI update that affects lease
> allocation, and the updates don't happen on both servers at the same
> time?

We've not used OMAPI.

> Also related: if we have an extended outage and I put the server
> into partner-down, can I continue to update the config on the
> working server, or should I stop the updates? Or do I need to
> guarantee in some unspecified fashion that the dead server can't
> re-animate until it has the newest config?

In this situation, I would probably try to make sure the configs are
synced if possible before restarting the repaired server, but if they
aren't for some reason I would imagine that things will catch up
eventually and just start working once the configs are synced.


More information about the dhcp-users mailing list