|
jmbyrdwell
9 posts
|
Howdy! Has anyone implemented GLBP for Load Balancing alonside BGP for incoming Redundancy? It seems by using the two together one would get the best of both worlds. I was thinking of each router receiving only a default route to its ISP (we have 2) and load balance across them using GLBP. Then use BGP for incoming redundancy. I'm curious if anyone has done this or has any thoughts as to why it may or may not work properly. Thanks! |
|
CaptnAmerica
24 posts
|
Was reading your post and thought it was interesting because it terms of loading balancing and routing protocols - they can do equal cost load sharing (eigrp of course you know it can use unequal cost load balacing through variance command). So the question is do you absolutely need to use one virtual IP to accomplish your goal? I'm not a BGP expert however I know BGP uses unicast, port 179 to start a neighbor session. What if those hellos were load balanced and your routes would flap? But do refer to this document about load sharing with two ISPs. http://www.cisco.com/en/US/tech/tk365/technologies_configuration_example09186a00800945bf.shtml#conf4 I know the BSCI does not mention ASA firewall with BGP because its out of the scope but hey who doesn't use ASA? ;-) Look at this document for BGP config with ASA in the mix - http://nextclickmedia.com/en/US/tech/tk365/technologies_configuration_example09186a008009487d.shtml |
|
jmbyrdwell
9 posts
|
Thanks for the response Captn. The reason for using GLBP and thus the Virtual IP is to provide a single gateway to my pix, thus eliminating any routing decisions on it's part. The routers would exchange BGP hellos beyond the scope of the GLBP. I've been using the first link you provided as a reference, but I don't really like the way they recommend "load sharing" when using two routers. I figure by using GLBP I can achieve more robust load-balancing. Here's what I have pictured:
----- -------- |PIX|-----| 2960 | ----- -------- | | --------- * ----| 2811 |------ ISP 2 --------- * |
|
CaptnAmerica
24 posts
|
Oh now it makes sense what you were trying to accomplish... :) And I thought the virtual IP was for another router forming BGP neighbors. Have never done BGP with GLBP but seems from your diagram it shouldn't have any problems. I can imagine after pix sends the packet to one of the routers through GLBP, the router will simply pick the best path for the destination. Anyhoo, are you picking some routes you want advertised from your ISP? I'm pretty sure you're not taking all the routes from your ISP with a 2800 series ISR. |
|
jmbyrdwell
9 posts
|
Oh yeah, I'm only accepting default routes from ISPs. Don't have the hardware for much more than that. |
|
CaptnAmerica
24 posts
|
If you are only accepting default routes, couldn't you just put static default routes on the PIX pointing to the two 2811 to load balance without using BGP? You can have your ISPs insert your route manually if you are looking to advertise your route. You don't have to be running BGP. |
|
jmbyrdwell
9 posts
|
Good call Captn. The more I think about it the less attractive GLBP seems. Someone advised me the asymmetrical routing that would result from GLBP would be a problem. What do you think? It seems to me that even with equal cost routing there would be asymmetrical traffic flows. I’d like to use BGP primarily for the redundant inbound internet connections. I’d rather not rely on manually configured ISP managed routers. |
|
CaptnAmerica
24 posts
|
In equal cost sharing without BGP, ISP A doesn't know that ISP B can also reach your network or BGP ASN number (no BGP). Therefore if the traffic was sent through Router A it would return to the router that has generated the traffic. Remember in BGP each hop is from AS to AS. Although you have two routers with differernt subnets, your BGP ASN is the same. This tells the ISP A and B and the internet that your ASN can be reached through either providers. So in BGP you can have asymmentrical traffic with the return paths. However with BGP you can use local preference for balancing exiting packets and then the med value for balancing incoming traffic with route maps. To verify, use extended ping and enable recording. Sort of like traceroute except this will also show you what IPs hitting during the return of the ICMP packet. :) Take a good look at this link again and look under verification. So with the route map statements, if the packet is sent out through router A, router A would be the best return path and so it is for router B. If you don't want to use interface tracking for GLBP weights BGP can help when one of your ISP goes does - BGP with next hop self command. Hope this helps a bit. Talking about Asymmetrical traffic flows it reminds me of this - crab |
Viewing 1 - 8 of 8
- 1
