DHCP server failover behind relay

[Vadym Abramchuk]

2012/9/5 Simon Hobson <dhcp1 at thehobsons.co.uk>

> Vadym Abramchuk wrote:
>
>  Unfortunally, none of my relay agents can send packets to both servers.
>>
>
> So the proper fix is to get the vendor of the broken relay agent to fix it
> ;-)


My local supplier said it's not going to happen. At least before Second
Marsian War ends ;-).


>
>  As for broadcast, I've tried setting the relay's server address to
>> 10.0.0.255 (broadcast of my servers subnet). dhcpd accepts DISCOVERs, but
>> the OFFERs do not reach the client.
>>
>
> Where do they get lost ?
> Also it might help to describe your network topology.


Thanks for your replies, Simon.

To simplify the things, here's a short cutoff: 10.0.0.70/24 is a primary
DHCP server (running at present), 10.0.0.90/24 is secondary one (actually,
it runs in virtual machine). Clients are at different subnet, 10.8.2.0/23.
There's a L3 switch routing between subnets (Dell 6024F in this case),
10.0.0.1/24 at one side and 10.8.2.1/23 at another side. The switch acts as
a DHCP relay.


>  Let's accept the axiom that I can't deliver requests to both servers. Is
>> there any other way?
>>
>
> Not that I know of.
>
>
What about IPVS (Linux IP Virtual Server) and/or VRRP? I could bind both
DHCP servers to same virtual server IP and switch between them with
external failover.
The only question is disabling all isc-dhcpd's internal failover features
while still syncing databases.

--
wbr,
Vadym Abramchuk
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20120906/de5ae5ab/attachment.html>