DHCP Failover

We are running two WIN2K3 DHCP servers using our cisco
routers as relay agents.

The scopes are split roughly 80/20 between two servers
with the 20% being reserved for failover.

Our intention was to have our leases obtained from the
primary DHCP server and only obtain leases from the
secondary if the primary is down. However, it appears
that both servers are leasing out IPs.

What is happening is that the DHCP requests are being
forwarded by the router to both servers at the same time
and the clients are obtaining leases from whichever
server responds first.

Is there any way to set up our DHCP infrastructure such
that our leases are obtained from the primary DHCP server
and will only obtain leases from the secondary if the
primary is down (third party app, reg hack, etc.)?

I know we cannot be the first enterprise that has run
into this.

Thanks!

Hi,

You may want to consider using DHCP on a MS Cluster server. See 325547
Support WebCast: Using WINS and DHCP on Microsoft Windows 2000 Clusters
http://support.microsoft.com/?id=325547


--

Thanks,
Marc Reynolds
Microsoft Technical Support

This posting is provided "AS IS" with no warranties, and confers no rights.


"JB" <(E-Mail Removed)> wrote in message
news:040201c3c428$0ca34b30$(E-Mail Removed)...
> We are running two WIN2K3 DHCP servers using our cisco
> routers as relay agents.
>
> The scopes are split roughly 80/20 between two servers
> with the 20% being reserved for failover.
>
> Our intention was to have our leases obtained from the
> primary DHCP server and only obtain leases from the
> secondary if the primary is down. However, it appears
> that both servers are leasing out IPs.
>
> What is happening is that the DHCP requests are being
> forwarded by the router to both servers at the same time
> and the clients are obtaining leases from whichever
> server responds first.
>
> Is there any way to set up our DHCP infrastructure such
> that our leases are obtained from the primary DHCP server
> and will only obtain leases from the secondary if the
> primary is down (third party app, reg hack, etc.)?
>
> I know we cannot be the first enterprise that has run
> into this.
>
> Thanks!
>
>

Hi!

I appreciate the response. However, I have been informed
that clustering is not an option at this point.

Are there are any other suggestions?

John Burtram


>-----Original Message-----
>Hi,
>
>You may want to consider using DHCP on a MS Cluster
server. See 325547
>Support WebCast: Using WINS and DHCP on Microsoft
Windows 2000 Clusters
>http://support.microsoft.com/?id=325547
>
>
>--
>
>Thanks,
>Marc Reynolds
>Microsoft Technical Support
>
>This posting is provided "AS IS" with no warranties, and
confers no rights.
>
>
>"JB" <(E-Mail Removed)> wrote in
message
>news:040201c3c428$0ca34b30$(E-Mail Removed)...
>> We are running two WIN2K3 DHCP servers using our cisco
>> routers as relay agents.
>>
>> The scopes are split roughly 80/20 between two servers
>> with the 20% being reserved for failover.
>>
>> Our intention was to have our leases obtained from the
>> primary DHCP server and only obtain leases from the
>> secondary if the primary is down. However, it appears
>> that both servers are leasing out IPs.
>>
>> What is happening is that the DHCP requests are being
>> forwarded by the router to both servers at the same
time
>> and the clients are obtaining leases from whichever
>> server responds first.
>>
>> Is there any way to set up our DHCP infrastructure such
>> that our leases are obtained from the primary DHCP
server
>> and will only obtain leases from the secondary if the
>> primary is down (third party app, reg hack, etc.)?
>>
>> I know we cannot be the first enterprise that has run
>> into this.
>>
>> Thanks!
>>
>>
>
>
>.
>

The catch here is, how would you determine that the
primary server is down? If it doesn't respond in a timely
manner? If so, "timely" would have to be defined
as "before the client request times out, leaving enough
time for the backup to respond, even if it's really
busy". I faced a similar problem many years ago on NT 4,
and decided there was no harm in having two dhcp server
assigning address as long as the scopes didn't overlap,
and neither server ran out of addresses (I only had about
a hundred clients so a class c split in two worked fine,
with about 32 addresses reserved for static devices).

A really ugly solution would be to use a service
monitoring tool (Logcaster, What's up, etc.) to monitor
the DHCP server on the primary server and start the DHCP
service on the backup machine when the primary failed. If
you are looking to provide "poor mans" redundancy, you
could use this to provide backup for other services that
don't have built in redundancy.

Good Luck!

>-----Original Message-----
>We are running two WIN2K3 DHCP servers using our cisco
>routers as relay agents.
>
>The scopes are split roughly 80/20 between two servers
>with the 20% being reserved for failover.
>
>Our intention was to have our leases obtained from the
>primary DHCP server and only obtain leases from the
>secondary if the primary is down. However, it appears
>that both servers are leasing out IPs.
>
>What is happening is that the DHCP requests are being
>forwarded by the router to both servers at the same time
>and the clients are obtaining leases from whichever
>server responds first.
>
>Is there any way to set up our DHCP infrastructure such
>that our leases are obtained from the primary DHCP server
>and will only obtain leases from the secondary if the
>primary is down (third party app, reg hack, etc.)?
>
>I know we cannot be the first enterprise that has run
>into this.
>
>Thanks!
>
>
>.
>

I sincerely appreciate your response.

We currently have both DHCP servers up with non-
overlapping scopes. The problem is that we are finding
that one server is being utilized more than the other.
What this results in, because of the 80%/20% split we
used, is one server is handing out all its IPs--in this
case the server with 20%.

Now, this is not impacting the normal function of DHCP to
the clients. However, looking at this point from a
disaster recovery point of view, we could see where if
the server with 80% of the IPs goes down and the server
with 20% is maxed out, we could begin seeing our clients
go down. Even with a 50/50 split, we see this could be
an issue for some offices as one server will respond
faster than the other server due to the number of
hops/latency/etc between servers.

Our goal is to have a solution whereby our clients are
not affected if either server goes down. We have about
3,000 clients spread out over the state on a routed WAN.
The logcaster tool sounds interesting, but I was hoping
there was some way to have a relay agent wait a few
seconds before forwarding the DHCP request on to the
secondary server.


>-----Original Message-----
>The catch here is, how would you determine that the
>primary server is down? If it doesn't respond in a
timely
>manner? If so, "timely" would have to be defined
>as "before the client request times out, leaving enough
>time for the backup to respond, even if it's really
>busy". I faced a similar problem many years ago on NT
4,
>and decided there was no harm in having two dhcp server
>assigning address as long as the scopes didn't overlap,
>and neither server ran out of addresses (I only had
about
>a hundred clients so a class c split in two worked fine,
>with about 32 addresses reserved for static devices).
>
>A really ugly solution would be to use a service
>monitoring tool (Logcaster, What's up, etc.) to monitor
>the DHCP server on the primary server and start the DHCP
>service on the backup machine when the primary failed.
If
>you are looking to provide "poor mans" redundancy, you
>could use this to provide backup for other services that
>don't have built in redundancy.
>
>Good Luck!
>
>>-----Original Message-----
>>We are running two WIN2K3 DHCP servers using our cisco
>>routers as relay agents.
>>
>>The scopes are split roughly 80/20 between two servers
>>with the 20% being reserved for failover.
>>
>>Our intention was to have our leases obtained from the
>>primary DHCP server and only obtain leases from the
>>secondary if the primary is down. However, it appears
>>that both servers are leasing out IPs.
>>
>>What is happening is that the DHCP requests are being
>>forwarded by the router to both servers at the same
time
>>and the clients are obtaining leases from whichever
>>server responds first.
>>
>>Is there any way to set up our DHCP infrastructure such
>>that our leases are obtained from the primary DHCP
server
>>and will only obtain leases from the secondary if the
>>primary is down (third party app, reg hack, etc.)?
>>
>>I know we cannot be the first enterprise that has run
>>into this.
>>
>>Thanks!
>>
>>
>>.
>>
>.
>