Massive L2/L3 topology problem

makemorebeer · 05-20-2008, 07:01 AM

I've got a real problem that seems to be growing worse by the hour. we have a main site and a remote site. they are connected via a cisco 1400 bridge and an ADSL link. there is the layer 2 and the layer 3. so my idea to get around this was to subnet off the remote site, and change all traffic to layer 3. here is where the problem comes in. I've got a router connected to a switch which is connected to a switch which is connected to another router. it's a strange topology so ill try and draw it out.

(ignor the periods. they are only for spacing.)

Core----router-------------switch
|........................................|
|........................................|
1400 wireless----router----switch

so in order to handle the routing I've got RIP setup ( I know but it's the only protocol my core switch will work with) on the core switch and the two routers. this part is working fine.

the problem comes in with the two switches. I've got the switches setup with Spanning-tree, and the routers have integrated switch ports which are also part of the spanning-tree with priority of 40(primary) on one and 50 (backup) on the other. this works sort of but the router needs to turn off completely in order for any type of fail over. a dropped link will not propagate through here.

is there any way to take a topology like this, and set it up so that if one link breaks it will auto fail over, or is this project doomed from the beginning.

interesting side notes. the core switch can handle ip routing, but not static metrics, hence the RIP usage. the other two switches can not handle ip routing so I'm trying to work around it with Spanning-tree.

TheWiz · 05-29-2008, 09:54 PM

I feel your pain. I have just posted a similar problem but I at least have switches that will both handle RIP. I assume you may have HP's as well as I found I couldn't use weighted static routes either. You seem to be suffering from the inability of a Layer 2 protocol to update a Layer 3 protocol of an outage. To solve your problem I would look at putting both of your remote site switches into the same Vlan connected to both the routers (assuming I have understood your diagram properly). I would then look at either VRRP on the LAN side of the routers (or HSRP if they are Cisco routers) and prioritise one as the primary gateway (most likely your ADSL link and router). You can also add metrics to learned RIP routes on link sides to affect traffic flows - but that is where my pain begins (see my thread).

Hope this has helped and its not too late for you.

Wiz

makemorebeer · 05-30-2008, 05:11 AM

I've actually reconfigured my topology =( I hate when that happens. I ended up using our backup coreswitch at the remote side which is capable of rip. and then I'm using EIGRP via the cisco routers to redistribute routes into the switch's Rip protocol. so far it has worked out fine in a lab environment. I'm about halfway through live implementation. so far it seems to be working. basically i've set the default route of the remote site devices to be the secondary coreswitch and then both the routers are connected to that as well as both remote site switches. you are correct, they are HP 5300, 2524, and 2824 switches. still have a single point of failure but it seems to be the best i can do at the moment. if only the original plan had worked. we had setup a wireless link via a third party company which would have kept everything layer 2 (STP and RSTP) but their techs couldn't figure out how to configure their devices so they would integrate into the Spanning-tree. i'll check into the protocols you listed though. maybe i'll find a jewel i can apply to my problem.

05-20-2008, 07:01 AM	#1 (permalink)
makemorebeer Registered User Join Date: May 2008 Posts: 7 OS: xp	Massive L2/L3 topology problem I've got a real problem that seems to be growing worse by the hour. we have a main site and a remote site. they are connected via a cisco 1400 bridge and an ADSL link. there is the layer 2 and the layer 3. so my idea to get around this was to subnet off the remote site, and change all traffic to layer 3. here is where the problem comes in. I've got a router connected to a switch which is connected to a switch which is connected to another router. it's a strange topology so ill try and draw it out. (ignor the periods. they are only for spacing.) Core----router-------------switch \|........................................\| \|........................................\| 1400 wireless----router----switch so in order to handle the routing I've got RIP setup ( I know but it's the only protocol my core switch will work with) on the core switch and the two routers. this part is working fine. the problem comes in with the two switches. I've got the switches setup with Spanning-tree, and the routers have integrated switch ports which are also part of the spanning-tree with priority of 40(primary) on one and 50 (backup) on the other. this works sort of but the router needs to turn off completely in order for any type of fail over. a dropped link will not propagate through here. is there any way to take a topology like this, and set it up so that if one link breaks it will auto fail over, or is this project doomed from the beginning. interesting side notes. the core switch can handle ip routing, but not static metrics, hence the RIP usage. the other two switches can not handle ip routing so I'm trying to work around it with Spanning-tree.

05-29-2008, 09:54 PM	#2 (permalink)
TheWiz Registered User Join Date: May 2008 Location: Sydney, Australia Posts: 84 OS: XP Pro SP3, Vista home premium (arghhh)	Re: Massive L2/L3 topology problem I feel your pain. I have just posted a similar problem but I at least have switches that will both handle RIP. I assume you may have HP's as well as I found I couldn't use weighted static routes either. You seem to be suffering from the inability of a Layer 2 protocol to update a Layer 3 protocol of an outage. To solve your problem I would look at putting both of your remote site switches into the same Vlan connected to both the routers (assuming I have understood your diagram properly). I would then look at either VRRP on the LAN side of the routers (or HSRP if they are Cisco routers) and prioritise one as the primary gateway (most likely your ADSL link and router). You can also add metrics to learned RIP routes on link sides to affect traffic flows - but that is where my pain begins (see my thread). Hope this has helped and its not too late for you. Wiz

05-30-2008, 05:11 AM	#3 (permalink)
makemorebeer Registered User Join Date: May 2008 Posts: 7 OS: xp	Re: Massive L2/L3 topology problem I've actually reconfigured my topology =( I hate when that happens. I ended up using our backup coreswitch at the remote side which is capable of rip. and then I'm using EIGRP via the cisco routers to redistribute routes into the switch's Rip protocol. so far it has worked out fine in a lab environment. I'm about halfway through live implementation. so far it seems to be working. basically i've set the default route of the remote site devices to be the secondary coreswitch and then both the routers are connected to that as well as both remote site switches. you are correct, they are HP 5300, 2524, and 2824 switches. still have a single point of failure but it seems to be the best i can do at the moment. if only the original plan had worked. we had setup a wireless link via a third party company which would have kept everything layer 2 (STP and RSTP) but their techs couldn't figure out how to configure their devices so they would integrate into the Spanning-tree. i'll check into the protocols you listed though. maybe i'll find a jewel i can apply to my problem.