Just a clarification. It is possible to have multiple subnets on an interface and configure the DHCP server to assign IPs to any of these scopes. No router address flip-flopping or other machinations are required or needed.
As has been posted, the primary IP address on the interface is *usually* (see details below) the giaddr placed into the DHCP packet by the relay agent (router). Lets assume that the interface doing the IP helpering has four subnets: P (the primary) and S1, S2, and S3. On any reasonable DHCP server, one can configure the secondary subnets to be "secondaries" to the primary in the DHCP config. So when one configures their DHCP server, they define the primary subnet information for P, and then define the information for S1, S2, and S3. One then ties these all together by making S1 a "secondary" of P on the DHCP server. Ditto for S2 and S3. The manner in which one makes S1, S2, and S3 secondaries is DHCP server dependent. If you have CNR and want to make S1 a secondary to P do the following: 1) Define the scope information for P, S1, S2, and S3. This would entail defining the range of address to hand out for each scope, the policy (DHCP attributes, selection tags, etc.) 2) Using the GUI, select the S1 subnet, then "properties" and then the "advanced" tab. Half way down, there a selection box to make this scope a secondary. Select this box, and when you do this, you can then select the primary for this scope. Select P. Note, this can also be done using the CLI. I believe the attribute name is "primary-scope" (or something close). Using the CLI, for scope S1, set its "primary-scope" attribute to the scope name you defined for subnet P. Once you;ve done this, when a packet arrives at the DHCP server with a giaddr of P, the DHCP server now knows that P and S1, S2, and S3 are all related. The DHCP server uses this, and any configurations the operator has provided to help select the appropriate scope (subnet) and thus IP for this device. Doing the above is very common practice in the cable industry. On any CMTS cable inteface, cable companies will have customer IPs subnets (for PCs) and subnets for cable modems. CPEs will be assigned globally routeable addresses (net24, net12, etc). and the cable modems will be assigned net10 addresses. The structure define above is used-- one of these subnets will be the primary on the CMTS interface and the rest will be secondaries. All are tied together on the DHCP server via the "priamry-secondary" logic described above. Cable operators configure the DHCP server logic to identify a DHCP request from a modem and map it to one of the subnet(s) on the interface created for modems. Ditto for PCs. Note, above I indicated that the primary address is *usually* the giaddr. Two caveats to this: * Cisco changed how the relay helpering works in some IOS revs-- in some 11.x or 12.x revs, the giaddr can cycle through all gateway addresses assigned on the helpering interface. That is, when a packet gets helpered, the router will initially insert the P address as the giaddr. If the DHCP server does not respond, and the router has helpered 3-4 DISCOVERs on behalf of a source, the 5-8th DHCP DISCOVER packets will get helpered using a giaddr of S1. This repeats 3-4 times, and if no DHCP response is received, S2 is used as the next giaddr. Note, the router maintains the state for each source so a new device will get helpered initially with P as the giaddr. (I don't recall when cisco enabled this cycling feature to be the defualt behavior. I believe they changed the default behavior back to only using "P" as the giaddr (I don't recall the IOS rev). However, I believe they've added a new know so that one can enable this "cycling" "feature" in current IOS revs. * On cable infrastructure gear (CMTSs), there are extra knobs to customize what value is inserted into the giaddr. One can configure the CMTS to always use the "P" address as the giaddr or to perform the cycling (described above). Michael Williams wrote: > > Plus, upon re-reading your post, I don't see an IP helper setup on the eth0 > interface on the spoke router just like you have on the hub router. You > need to add that. > > The point of my previous post was to highlight the fact that you need to > make sure that the primary IP on the eth0 on the spoke router be in the same > subnet with the IPs you want to hand out via DHCP. AFAIK, it's not possible > to service multiple subnets simultaneously on a single interface via > IP-Helper. (i.e. I don't think it's possible service any secondary IP > subnets on eth0 at the spoke site because the IP-Helper uses the primary > eth0 IP as the source address for the DHCP directed request) > > Mike W. Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=35602&t=35539 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
