Router Redundancy Questions

[Jkloza_2]

Looking for a recommendation.  Question about load balancing between point to point WAN links. 

I'm currently looking for a solution to provide more network redundancy over my two point to point DS3 lines
that connect my remote site.  I have 2 Cisco 3845 routers that terminate a point to point DS3 between two locations.  We are standing
up a second DS3 between the sites, and have purchased an additonal 2 routers.  I would like to either setup load balancing / sharing, or HSRP / VRRP between the routers / sites.

I would rather be able to utilize both lines, and if there is an issue with one, have traffic seemlessly transition to the other active line.  What would be the best way to accomplish this.  I'm in the process of also converting the sites from static routes to EIGRP.

Thanks, and if more information is needed, please let me know.

Jon

[Jon Marshall]

Jon

On the WAN side just connect them up. Whether it load-balances is to do with what you do on the LAN side. If you use HSRP on the LAN interfaces then it won;t load-balance.

It depends on what is behind the WAN routers in each site. If you have a L3 switch(s) then the easiest thing to do is connect the routers back to the switch with 2 P2P links using /30 subnets.

Can you be more specific about what the routers connect back to at each site and then we can be a bit more precise.

Jon

[Jkloza_2]

Sorry, yes, I can offer more specifics. 

At my data center, I connect to a L3 switch using a /30 IP subnet.  At my remote site, my router does all routing of VLANs for the network via sub-interfaces, which then connects to a L2 / L3 switch. 

It can easily be changed to a L3 switch, I have a 3750 that I'm using right now as just a L2, but I can upgrade the IOS, and turn on IP routing, because it's a fairly small site, with a handful of hosts (less than a /24).

That said, I'm not sure that I understand what you mean by the load-balancing will be done on the LAN side.  My current DS3 is sitting around 60% utilization, so I'd like my second connection that I'm bringing up to be used as well, not just be a redundant line, only passing traffic if there's an outage on the primary line. 

I can provide a network drawing to show how it's currently connected, and what I'm trying to do if this isn't getting the point across.

But now I have - L3 Switch (6509E) -----------------> 3845---------DS3------------3845 ---------> L2 Switch - 3750

What i'm looking to do is add 2 more 3845's, and a new DS3 to offer more network redundancy to my current infrastructure, but to also offer some additional bandwidth to the users at my remote site.

Thanks,

Jon

[Jon Marshall · Accepted solution]

I meant there is nothing special to do on the WAN side in terms of config - just connect it up. The config needs to be done on the LAN side, apologies for the confusion.

If the WAN routers are to connect to the same L3 switch at each site then by far the easiest thing to do is to connect both WAN routers to the switch using 2 x /30 P2P links and use a dynamic routing protocol to exchange routes between the WAN routers and their corresponding switch.

Each L3 switch should then see 2 equal cost paths to the remote site and will do per destination load-sharing.

At the remote side if the WAN routers actually do the routing for the clients directly ie. no L3 switch you could use GLBP on the LAN interfaces of the WAN routers and this would load-balance from the clients. Or as you say you could make it a L3 switch and just have the same config in both sites.

Note - i am assuming that you will be running a dynamic routing protocol across the WAN and back to the L3 switch(es). You could just as easily do it with static routes but then you would need to look at IP SLA in case one link fails. A dynamic routing protocol would take care of this for you.

Jon

[Jkloza_2]

Jon,

Understood.  Thank you very much.  I can convert my switch @ my remote site to L3, it will take some of the load off my routers.  Also, we'll probably be putting in a 4507E pretty soon anyway, so it's not an issue. 

Currently I'm using static routes between sites, and in my core, but I'm planning on converting everything to EIGRP very soon, as the static routing tables are getting to be a bit much to manage.  It just makes more sense to do this now, then to wait, but once I've converted to EIGRP, I'll then bring up the redundant connections.

For the equal cost paths, my question would be how is the best path chosen?  I mean, it's technically just two point to point lines, as long as everything is up / up, which path is chosen? 

Also, if I experience problems with one of the WAN lines, would this take send traffic back across the line that is still up?

Thanks again for all of your help!

Jon

[simon.dwyer · Accepted solution]

The path is chosen using CEF.

If there are two equal routes in the routing table the router/switch will use CEF to choose which one to use. 

This will explain it better than i can off the top of my head

http://www.cisco.com/en/US/products/hw/modules/ps2033/prod_technical_reference09186a00800afeb7.html

[Jon Marshall · Accepted solution]

Jon

As already noted by Simon it will use CEF per destination load-balancing.

To answer your other question. If one of the links goes down then the routes will be lost on that link and then the switch behind the router will not receive routes from that WAN router. This is where dynamic routing protocols excel.

If, as i mentioned before, you used statics then you would need to use something like IP SLA and this complicates things so going with EIGRP would be a good move.

Jon

[Jkloza_2]

All,

Thank you so much.  Should I post in a new topic about converting to EIGRP routes from static, or is it OK to continue this topic?

I was just wondering if there's a way to advertise all directly connected routes in my L3 switches / routers in EIGRP, or if it's better to advertise the networks seperatley?  Also, this should not affect my network right?  If setup properly, I should be able to implement EIGRP, and pull my static routes out one by one, once I see that everything is advertised, and the neighbor relationships are formed properly in the EIGRP topology.

Thanks!

Jon

[Jon Marshall]

Jon

You can advertise them together if you want. So if you had for example networks -

192.168.1.0 -> 192.168.10.0

you could have individual entries

router eigrp 10

network 192.168.1.0 0.0.0.255

network 192.168.2.0 0.0.0.255

.....

network 192.168.10.0 0.0.0.255

no auto-summary   <-- make sure you add this

or you could do -

router eigrp 10

network 192.168.0.0 0.0.15.255

no auto-summary

note 192.168.0.0 0.0.0.15.255 would actually cover 192.168.0 -> 192.168.15 subnets. You could if you wanted to be more specific do -

router eigrp 10

network 192.168.0.0 0.0.7.255

network 192.168.8.0 0.0.0.255

network 192.168.9.0 0.0.0.255

network 192.168.10.0 0.0.0.255

no auto-summary

or if you are uncomfortable with wildcard masks -

router eigrp 10

network 192.168.0.0 0.0.255.255 <-- which covers subnet 192.168.0 -> 192.168.254.0

no auto-summary

i'm not trying to confuse you, rather the point i am trying to make is it is all about control ie.do you want be specific or are you happy to be not so specific. Note all the above will still only populate the routing tables with your actual subnets ie. 192.168.0.0 0.0.255.255 wouldn't actually advertise out all 255 subnets, it would still only advertise 192.168.1.0 -> 192.168.10.0. It just means if you later add another 19.168.x.0/24 subnet with 192.168.0.0/16 it would automatically be advertised which may or may not be what you want.

As for setting it up. Yes you should be able to do this with no downtime in terms of populating EIGRP routes. You won't see them in the routing table because the statics will be there. To make sure the router has received them, run this command -

sh ip eigrp topology all-links

Jon

[Jkloza_2]

Jon

I understand how I can advertise my networks via wildcard mask.  I think i mis-worded my question.  I meant if i used a "network 0.0.0.0", or something along those lines, will it auto-advertise all of the directly connected routes in the routers routing table?

Sorry for the confusion, and I appreciate all of the help.

Thanks,

Jon

[Jon Marshall]

Jon

Apologies if i was telling you something you already knew.

Yes you could do that as well although i would generally use the actual common octets for the subnets but it would work.

It's important to realise with the network statements under EIGRP that you are actually telling the router which interfaces to start EIGRP on rather than simply which networks to advertise. So that is sometimes why you want to be more specific.

Jon

[Jkloza_2]

Jon,

Thank you very much for all of your help.  I just finished setting up EIGRP and pulling out my static routes, everything worked like a charm. 

I'm standing up the redundant line today, but we're having carrier problems (big suprise ). 

I just have one last question about the load balancing.  In my L3 switch at my remote site i have a default route pointing to the directly connected DS3 router.  After standing up the new router & connecting it back into my network as described above won't something have to be configured so that the traffic taking the default route is load balanced between the two circuits?

Thanks again for all of your help!

Jon

[Jon Marshall]

Glad to hear things are going well.

The remote site switch, is it L3 and are you going to run EIGRP between that switch and the 2 routers ?

If so EIGRP will take care of all the internal networks you have. If you still need a default-route you can simply add a 2nd default-route to the swiitch pointing to the new router with the new line and this would work fine.

If you had multiple sites a better solution would be to configure the default-route at HQ and then redistribute this into EIGRP. Then all your remote sites would get the default-route from both WAN routers. But if you only have one remote site it is simple just to add it to the switch.

Note though that if all traffic that is not local to the remote site only goes to HQ then you would only need these static default-routes ie. the EIGRP routes would be superfluous.

Having said that you wanted to run EIGRP and this is more scalable for the future in terms of additional sites. So i would add your default-route at HQ and then redistribute into EIGRP. You would then remove the static route from the switch and allow the 2 remote site WAN routers to pass that back to your remote site switch.

One last thing.I'm assuming the only entry point into your remote site is the WAN and it used to get to and from HQ ie. there are no other connections at the remote site that go elsewhere. If that is the case then you may want to look into EIGRP stub configuration if/when you add any further sites. But it is not that important now.

Any futher questions etc. don't hesitate to ask.

** Edit - technically if all remote traffic from your remote site goes out via the WAN you only actually need a default-route in the remote sites. So you may actually want to look into EIGRP stub now. What EIGRP stub will allow you to do is only accept the default-route and then advertise out the local subnets. So you don't need all the HQ subnets in the routing tables at the remote site as the default-route will take care of the traffic from the remote site to HQ.

Jon

[Jkloza_2]

Jon,

Wow, so many different options.  Let me see if I can answer your questions without missing something.  The remote side switch is L3 now, and yes it is running EIGRP between the routers.  Unless i'm missing something, I need default routes so that my systems at the remote site can get to the internet.

Right now we have a few static routes pointing to another location that hangs off of my remote site.  I didn't include this information because the site is going away in less than 15 days, and I didn't even bother bringing it into the new EIGRP topology. 

It doesn't matter to me hoow I accomplish this configuration, if 2 default routes in the L3 switch at my remote site will do the job, thats fine.  I've never had 2 default routes in a device before, I actually thought that you could only have one.  The EIGRP stub configuration sounds interesting, in 15 days, all traffic at my remote site will be sent back to HQ.

When I advertise a default route in EIGRP at my HQ site, any router that I have configured in my EIGRP topology will get the default route right? 

Thanks for all of your help.

Jon

[Jon Marshall]

Jon

Yes, sorry i didn't want to confuse you

Yes a L3 device can use more than one default-route, it will simply alternate between each route usually using per-destination load sharing.

Yes if you add the default-route and redistribute into EIGRP then all EIGRP speaking routers/L3 switches will see it.

You are not missing anything, you do need default-routes at the remote site. If you have another location hanging off the remote site do not use EIGRP stub.

If you are happy for all L3 devices to receive the default-route then i would redistribute it from EIGRP at the HQ site and remove the static default-route off the L3 remote site switch. Once the other site has been removed, remove the statics pointing to that site from the L3 switch and possibly look into EIGRP stub.

One last point. If HQ needs to get to that small site hanging off your remote site you will need to temporarily redistribute these into EIGRP so that HQ knows how to get to them.

Jon