You can do this, yes.

Your local hosts would point to the GLBP VIP that both R1 and R2 share.

The problem arises when one of the routers loses its connection to the service provider network. You need to leverage the weighting and object tracking features to make sure that a router is no longer advertised as an AVF once it loses its connection to the WAN.

http://www.cisco.com/en/US/docs/ios/12_2t/12_2t15/feature/guide/ft_glbp.html#wp1046944

HTH

Victor

Good catch, Joseph.

I looked at the drawing too quickly and didnt realize that it was a L3 switch between the LAN and the WAN routers.

Rated.

Victor

Ok...L3 is between the hosts & both routers..L3 is planned to act as g.w for hosts & this switch will den point to vip of glbp for both routers...or simply the interfaces of L3 getting the upstream routers links would be enabled with ospf to achieve outbound loadsharing.Which one should be better ..merits & demerits?

also will glbp have arp issues in this setup?

Thanks.

You do not want to use GLBP between routers (and/or L3 switches running L3). Reason being, GLBP will usually only "see" one source MAC (i.e. there would be no load balancing).

You can use MHSRP and static routing, or dynamic routing. Either works, which to choose depends on other factors. (Normally, I learn toward dynamic routing.)

You mean the mac seen will be only the one from the switch until the timeout occurs..so it wont loadshare with this setup?Pls throw some more light to get me understand this.

& how if i enable ospf on 2 interfaces of L3 as routed ones and so will it loadshare?

Thanks.

If you statically route from an L3 switch to GLBP addresses, GLBP will only see the source MAC for the L3 switch. I.e. as far as GLBP is concerned, all flows are from one host, so all traffic will use just one gateway.

OSPF load share should (usually) be automatic once the L3 switch sets up an OSPF adjacency with both routers assuming cost to destination is the same via both routers.

You configure OSPF to match the network or networks shared between the L3 switch and the routers. Depending on the L3 switch, the L3 switch ports that connect to the routers might be routed ports or switch VLAN ports. If VLAN ports, they could be VLANs unique to each router, used by both routers, used by both routers and other hosts, or other mixture of hosts and routers.

PS:

You might want to avoid having other hosts share your external router subnets, but if they do, you'll also want to avoid the L3 switch allowing redirects.

That educated me!so as the switch would arp here..it wont be ideally useful.

So what scenario does glbp fit nicely in doing its task correctly..

Thanks.

GLBP is typically used as a first-hop router redundancy protocol, like VRRP or HSRP, only GLBP leverages load balancing to take advantage of the available bandwidth.

In a LAN environment, there will be many different users who will ARP for the default gateway's MAC address. The AVG (Active Virtual Gateway) router that controls the forwarding group will return to each user one of the AVF (Active Virtual Forwarder) MAC addresses. Hence, the load balancing.

HTH

Victor

"So what scenario does glbp fit nicely in doing its task correctly"

Lets say you have an access-layer switch (AS1) that is connected to 2 distribution switches (DS1 & DS2).

The inter-vlan routing for the clients connected to AS1 is handled by DS1 & DS2. AS1 is connected to DS1 with a L2 trunk and AS1 is connected to DS2 with a L2 trunk.

DS1 & DS2 are connected together via a L3 connection. Because there is no L2 loop AS1 can use both connections to forward traffic from clients ie. either AS1 -> DS1 or AS1 -> DS2.

GLBP works well here as some clients will use one link because DS1 is their default-gateway and other clients will use the other link because DS2 is their default-gateway.

Note that if the link between DS1 & DS2 is a L2 trunk GLBP still works but not as well because one of the uplinks from AS1 will blocked so even though the clients would still use both DS1 & DS2 as default-gateways all traffic would go up one link and then some would have to go across the DS1 -> DS2 link.

You could manipulate the STP costs so the L2 trunk between DS1 & DS2 is blocking but thsi may affect other traffic flows in DS1/DS2.

Jon

Jon:

Good morning/afternoon ( for you):

Not to be nit-picky, but just to be crystal clear for the OPs sake. When you say:

"GLBP works well here as some clients will use one link because DS1 is their default-gateway and other clients will use the other link because DS2 is their default-gateway."

It may be worth mentioning that all hosts will have the same default gateway IP address configured, and that will be the VIP for the GLBP group. What will be different in each host is the IP-to-MAC mapping. The AVG will assign each user an AVF in the group to use to forward their traffic to.

I am 110% confident than you know this, it is just for the OP's sake that I am getting anal. I thought it was worth the effort...

Thanks

Victor

Victor

Morning for you.

"It may be worth mentioning that all hosts will have the same default gateway IP address configured, and that will be the VIP for the GLBP group"

Not nit-picky at all my friend. Well worth pointing out as the other way of load-balancing ie. MHSRP does indeed use different default-gateways on the clients.

Jon

Jon, I really like this explanation because of the way you tied GLBP with STP. I think you made an important contribution to the discussion that was missing.

Rated.

Victor