[Dnsmasq-discuss] Redundant setup with Dnsmasq

[AJ Weber]

Or, I think you could skip that if you setup the two, "actual" servers 
to NAT all responses appropriate to DNS/DHCP ports, so everything always 
looks like it's responding from the .250 address, regardless of whether 
the .251 or .252 server actually sent the response.

I could be wrong.


On 12/22/2011 11:44 AM, Jan Seiffert wrote:
> 2011/12/22 Markus Schöpflin<markus.schoepflin at comsoft.aero>:
>> Thank you for your idea. This really seems OK for our needs. If I understand
>> things correctly, I would have to do that on all four LANs the current Dnsmasq
>> is serving. Just one small additional question:
>>
>> Am 22.12.2011 15:13, schrieb Michael Rack:
>>
>>> Very easy.
>>>
>>> You need at least one virtual ip-address for your DNS- and DHCP-Server.
>>>
>>> So lets say you have a Class-C Network 10.0.0.0/24
>>>
>>>        * Primary DNS / DHCP    10.0.0.251
>>>        * Secondary DNS / DHCP  10.0.0.252
>>>
>>> Now, you add a virtual IP to your primary DNS - lets say
>>>
>>>        * Virtual-IP            10.0.0.250
>>>
>>>    From Secondary you create a Bash-Script that do the following:
>>>
>>>        * Check the Server-Status by ping the virtual ip-address
>>>        * when the ping has failed:
>>>           * add the virtual ip-address to your network-configuration
>> Wouldn't it make sense to send an unsolicited ARP packet to update the ARP
>> caches of neighbours after the IP address has moved?
>>
> Yes.
> I was about to write the same tip.
> Sometimes ARP-Tables can have a quite long timeout, so the "failover"
> would be stuck.
> Maybe something along the lines of
> arping -c 1 -A -s 10.0.0.250 $BROADCAST_ADDR
>
> [snip]
>> Thank you,
>> Markus
>>
> Greetings
> Jan
>