DHCP Configuration on Redundant Controllers

[mitel-user]

We are using two controllers as DHCP servers at two different locations.  The network side is configured properly (ip helpers, etc.).  We would like to use the exact same IP address ranges on both DHCP servers so in case of failure of the primary, the secondary would kick in with no visible IP address changes.  I know this is completely normal configuration for other DHCP servers (Windows, Linux, etc.) but I have been told that Mitel does not support this well and we are risking running into a situation where both controllers give away the same IP address causing duplicate IP address clashes.  Can someone please confirm this?

[ralph]

The 3300 would have be able to sync IP addresses with each other.  Doesn't do it.

Ralph

[bobcheese]

Curious as to why the Il address matters if you are using DHCP?

[mitel-user]

Ralph, I don't quite understand what you are saying.  Can you please expand your answer a little bit if possible?  Thanks.

[Mattmayn]

I'm with Bobcheese on this one. Is there a specific reason why you need the IP addresses to match?

[ralph]

I'd also have to question why the IP address would make a difference.

But, to answer your question about the sync -  If you have two servers handing out IP addresses from the same scope and range, they have to be aware of what addresses the other one has already handed out so you don't get dupe IP addresses.   A dupe address could crash a device - not a good thing.   That would be a sync function the 3300 is not designed to do.

But, I guess the first question would have to be: why do you need to do this?

Ralph

[bobcheese]

Also curious as to why u would want 2 3300's running on the same subnet, surly it would be better to have different subnets. If you do need this setup then why not use a single DHCP source like win server?

[Mitel100]

Normally when you configure DHCP for redundancy, you would normally do a 80/20 split across the two DHCP servers, where the primary server has 80% of the address and the secondary has 20%. The DHCP servers would never reference the whole subnet, this is not normal configuration practice.

[bobcheese]

Mitel100, can you confirm how this would work then? presume option 125 would contain multiple controller addresses as in most cases primary controller houses all IP sets and secondary is purley a resilient controller. Also if this was the case and only 20% scope on secondary how do you cover off the scenario where the primary fails and then (for some reason power issue maybe) the phones reboot and broadcast for DHCP? Now with only 20% on secondary (presuming you have a 100+ users) there is not enough IP addresses to service all sets?

Excuse my ignorance but I have had limited experience to resilient setups.

[mitel-user]

Thank you all for the replies.  I think some of the people misunderstood my question.  The two controllers are not on the same network, they are not even in the same geographic region.  They just happen to hand out the same IP space to IP phones, one being the primary and the other a secondary DHCP server which kicks in if and only if the primary fails.  The two IP DHCP pools must be the same size so that they can accommodate all users in a call center.  Also these pools have to match the IP addressing of the network interface, which cannot be changed (part of a longer conversation).  Also, the way a DHCP server knows which IP addresses have been assigned by the another DHCP server is via the DHCP mechanism of broadcasting requests and responses.  This is where my question comes in.  The closest answer was from user Mitel100 who recommended we split the space 80/20.  My question again is can we do this with the exact same IP ranges, why or why not?

[Mitel100]

Hi Bobcheese, the split really depends on the environment, the 80/20 split was an example of a split scope that is typically used when two DHCP servers are located on the network. The DHCP services will normally work in tandem, with no one service being the primary or secondary.

The DHCP scopes will normally be planned so that if you do have 200 agents, then a /23 network can be introduced to cater for the split.

Mitel-user, Mitel are correct in saying this set-up is not really done well on the MCD, as it is only ment to be used for small environments. the problem with having two DHCP servers running the same addresses is that when a DHCP request is sent back to the DHCP servers, based on the Transaction ID field in the request, servers are informed whose offer the client has accepted. When other DHCP servers receive this message, they withdraw any offers that they might have made to the client and return the offered address to the pool of available addresses, so the same IP address becomes available on the 2nd DHCP server for re-distribution.

However, typically, a client may use the Address Resolution Protocol (ARP) to prevent IP conflicts caused by overlapping address pools of DHCP servers, but I don't think the IP phones have this.

You may have to look at increasing the scope.

It may help to know how many end devices we are dealing with here and the subnet class, so I can best help you with a configuration.

[mitel-user]

Our needs are for about 200 phones/users per building floor with maximum potential growth to 220.  We have assigned /24 subnets per floor which should be enough if it wasn't for the backup requirement.    It sounds like with the secondary DHCP server we will need a /23 and assign the lower /24 pool to the primary and the higher /24 to the secondary.  This also implies changes to the network gateway interface IP address mask and assumes contiguous IP subnet space is available.  Is this correct?