I think the picture got messed up. But, let's say R1 and R2 are running HSRP on the Local LAN. It doesn't matter which one becomes primary. If the clients send to one router, but the other router has a better route, than the router will send the packet back out the Local LAN to the other router. It's the typical extra hop that many networks have. The router should send an ICMP Redirect (although that is disabled by default when using HSRP.) But it works without any major hitches because both routers have complete routing tables that describe the entire internetwork.
Since your picture is symmetrical (or at least I think it was?) the same thing can occur on the Corporate LAN. R3 and R3 can run HSRP too. Now, for traffic coming back, we have a more interesting problem.... It would depend on the routing protocol and the maximum-paths configuration, wouldn't it? For some routing protocols, each router would only know one way back. If that way includes the broken interface, then the protocol will have to converge before traffic can make it back. A few more comments in line... >Consider the following: > > Local_LAN > | > -------------------------------------- > | | > R1 >R2 > | >| > telco_1 >telco_2 > | >| > R3 >R4 > | >| > ---------------------------------------------- >-------- > Corporate_Network > > >Seems to me that of R3 and R4, the coproarate network knows one of those as >the route to the Local_LAN, preferably the router that is the HSRP primary. You mean the HSRP primary on the Local LAN? Of course the routers on the Corporate Network don't know anything about HSRP on the Local LAN. Plus, it doesn't matter whether their path goes back via R1 or R2. Which one it chooses would depend on the routing protocol. Maybe it's IGRP and one of the links has much less bandwidth so the other is preferred. Maybe you're using variance so that both routes are known. >hhhmmmm........ thinking about this, interesting design study. HSRP effects >only Local_LAN traffic to the Corporate_net. Does return traffic route >matter? HSRP on the Local LAN doesn't affect it. Other things do. >hhmmmmm..... would good design consider that R3 and R4 also be an HSRP pair? In your simple design, sure, I would say make them HSRP pairs too. You might want to know some load balancing and make one the active for some VLANs and the other the active for other VLANs. I know you know all this basic stuff. ;-) If you meant for this to be a more advanced discussion, just let me know. Thanks. Priscilla >If they were, what would the effect be, as opposed to if they were not >Maybe I'm outsmarting myself about the data flow implications? > > > > > > Certainly, one can create scenarios where load-sharing or other > > factors make the two routers significantly different. Depending on > > the goals and budget, you might even have HSRP in edge routers and > > more complex routing at a distribution tier. > > > > For that matter, people often don't consider L2 failover techniques > > (e.g., UplinkFast and EtherChannel) with switches feeding the HSRP > > routers as another aspect of no-single-point-of-failure. > > > > > > > >----- Original Message ----- > > >From: "Howard C. Berkowitz" > > >To: > > >Sent: 23 June 2002 3:54 pm > > >Subject: Re: Re: HSRP [7:47177] > > > > > > > > >> At 3:08 PM -0400 6/23/02, Kevin Cullimore wrote: > > >> >A useful notion to keep in mind is that hsrp and its un-patented > > >> >counterparts (you'd think that during the past century, people would > > >learn > > >> >from IBM's example, but apparently that isn't the case) are >profoundly > > >> >asymmetric in scope: > > >> > > > >> >they are concerned with the host->default gateway portion of the > > >> >conversation, not the return path (although implementational >specifics > > >might > > >> >force them to address the return path in some circumstances). > > >> > > >> > > >> Kevin, how is the asymmetry a problem? The HSRP linked routers > > >> presumably have the same routing tables, although the backup might > > >> have to ARP for its first packet forwarded. Even if that's an issue, > > >> promiscuous ARP learning shouldn't be all that much of a problem. ________________________ Priscilla Oppenheimer http://www.priscilla.com Message Posted at: http://www.groupstudy.com/form/read.php?f=7&i=47300&t=47177 -------------------------------------------------- FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
