DHCP fail over address delay

[Glenn Satchell]

That's certainly odd behaviour. However we probably need a little more 
information to help you.

What version of dhcpd are you running (dhcpd --version) ? What Os 
version, and what hardware platform (eg x86, sparc, hp, etc)

Is there anything in the logs? What state does dhcpd start in? Is it 
"normal", "communications-interrupted" or something else? What log 
messages are there on both servers when the request is failed over to 
server2?

What is the network layout like? Are server1 and server2 on the same 
subnet, in the same data centre or geographically far apart?

regards,
-glenn

On 04/23/10 08:16, Bryan Dobson wrote:
> Having a strange issue regarding DHCP fail over which I was wondering
> anyone else had some across in the past. I have gone over the
> configuration and believe everything is set as it should be. The issue
> is we have a pair of DHCP servers set in a fail over configuration and
> have been seeing excessive delays in IP addresses behind handed out.
>
> The servers are running the same version of DHCP and the time on both
> servers is in sync. If the request comes into server1 and it does not
> fail over, it will complete the transaction and the user will get an
> IP. However, if the request is failed over to server2 I have seen a
> delay of sometimes more than an hour.
>
> On server1 we see repeated DHCPDISCOVER to peer until eventually
> server1 provides the ACK for the address itself.
>
> server1 failover:
>
> failover peer "dhcp" {
>    primary;
>    address server1;
>    port 51000;
>    peer address server2;
>    peer port 51000;
>    max-response-delay 60;
>    max-unacked-updates 10;
>    load balance max seconds 3;
>    mclt 7200;
>    split 128;
> }
>
> server2 failover:
>
> failover peer "dhcp" {
>    secondary;
>    address server2;
>    port 51000;
>    peer address server1;
>    peer port 51000;
>    max-response-delay 60;
>    max-unacked-updates 10;
>    load balance max seconds 3;
> }