ip summary-route (in RIP and EIGRP)

[SJ K]

Hi all,

On interfaces with ip summary-address set;

RIP

1) can we use ip summary-address for RIPv1 ? or v2 or both

2) when we issue ip summary-address, will the interface still send out full route tables updates periodically (e.g. 30 sec) ?

3) if not, will it still send out the summary route periodically ? or its just a 1 time basis

4) if there are any internal routing changes, will it impact the interface to send out any changes for the summary route ?

5) what are the scenarios whereby the summary route will be removed by the adjacent router (e.g. when it no longer hear the summary route update from the sending router ?)

EIGRP

1) when we issue ip summary-address, will the interface send out summay route periodically ? or its just a 1 time basis

2) if there are any internal routing changes, will it impact the interface to send out any changes for the summary route ?

3) what are the scenarios whereby the summary route will be removed by the adjacent router (e.g. when it no longer hear any hello from the sending router" ) ?

Regards,
Noob

[Bilal Nawaz · Accepted solution]

RIP

it doesnt make sense using summary-address for RIPv1 most of the time because its a classful protocol. You could but I think same rules would apply as v2. v2 is classless and makes more sense to use the summary here. For RIP there is a possible 2 ways to summarise. 1 - automatically by summarizing the subprefix's to the classful boundry with the auto-summary command. 2 - a specified summary address configured at some other classless / classful boundary. Similar to EIGRP too i guess.

When RIP determines that a summary address is required in the RIP database, a summary entry is created in the RIP routing database. As long as there are child routes for a summary address, the address remains in the routing database. When the last child route is removed, the summary entry also is removed from the database. This method of handling database entries reduces the number of entries in the database because each child route is not listed in an entry, and the aggregate entry itself is removed when there are no longer any valid child routes for it.

RIP Version 2 route summarization requires that the lowest metric of the "best route" of an aggregated entry, or the lowest metric of all current child routes, be advertised. The best metric for aggregated summarized routes is calculated at route initialization or when there are metric modifications of specific routes at advertisement time, and not at the time the aggregated routes are advertised.

The ip summary-address rip routerconfiguration command causes the router to summarize a given set of routes learned via RIP Version 2 or redistributed into RIP Version 2. Host routes are especially applicable for summarization.

The updates are the same as usual, every 30sec's by default.

EIGRP

EIGRP sends partial updates only when it needs to and only for the prefix's it needs to update. So when we summarise at the time of hitting that "enter" key, its a one time thing, unless something changes it will remain in the topology as long as their is a child route that exists in the RIB/topology.

As long as there is at least one child route that match's the parent summary then there should be no impact.

The scenario where the summary route will be removed is if the advertising router does not have a child route in its RIB, and thus withdraws the route from being advertised.

Of course I hope i'm not wrong but this is how I always envisaged the summarisation to work. Maybe the super experts like Peter and Jon can explain better and correct me if im wrong :)

hth

Bilal

[SJ K]

Hi Bilal,

Thanks! I think we can forget about RIPV2 all together. It doesn't allow supernetting in its summarization.whereas EIGRP can.

Thanks!

[Peter Paluch]

Hi Bilal,

Very nice answer! Regarding your comment about the "super expert" - you're exaggerating a bunch, you know that? :)

it doesnt make sense using summary-address for RIPv1 most of the time because its a classful protocol

Actually, in RIPv1, manual summarization should not be possible at all. Doing summarization inside the same major network, e.g., aggregate 10.0.0.0/24 - 10.0.3.0/24 into 10.0.0.0/22 and send it out over another interface in the 10.x.x.x network would break the assumption that all subnets of the major network have the same netmask that can be inferred from the interface itself. After all, RIPv1 messages do not have the netmask field so it would not be even possible to advertise the summary mask at all. And obviously, when advertising subnets of one major network out an interface in a different major network, the automatic summarization kicks in by itself, as that is a required RIPv1 behavior. Doing a different summarization at a boundary between major networks would fail again because the receiving router would assume a classful mask - there's no way of telling it that the summary mask is different.

So to answer Koh's questions one by one:

RIP:

1) v2 only

2) The summary route will be sent every 30 seconds along with other networks that do not fall under the summary address and are advertised without change. RIP is a timer-driven protocol sending all known routes repeatedly even if nothing changes. Networks that fall under the summary address are not advertised, of course, only the summary address is advertised on their behalf.

3) N/A

4) The only change in the summary route would be its metric. As Bilal explained, in Cisco implementation, a summary address inherits the lowest metric from among the routes that were summarized. If this minimal metric changed, so would the metric of the summary.

5) A summary address will stop being advertised if there are no networks falling under that summary. The adjacent router does now know whether a learned route is a summary so there are no specific rules about learning or removing it.

EIGRP

1) After configuring a summary, it will be advertised only once. EIGRP is an event-driven protocol so if nothing changes, routes are not re-advertised.

2) The same as with RIP: if the smallest metric from the component routes changes, this change will propagate to the summary address and it will be advertised with an updated metric. Otherwise, nothing will be advertised.

3) The same as with RIP.

I think we can forget about RIPV2 all together. It doesn't allow supernetting in its summarization.whereas EIGRP can

Well, this is a limitation in Cisco's implementation, not in the design of RIPv2 itself. Just try creating a static route, say:

ip route 10.0.0.0 254.0.0.0 Null0

and then in RIP, configure the redistribute static command - and you'll see this network being advertised even though it is a supernet encompassing 10.0.0.0/8 and 11.0.0.0/8.

Best regards,
Peter

[SJ K]

Hi Peter, Bilal, Jon, and everyone,

Sorry for the late reply as I have been to 3 interviews on Friday and the interviews suck the soul out of me..

You guys are all the experts here.. I am the ultimate... noob.

======================================================

Back to the topic.. its Sunday, I hope Peter and you guys are still around tomorrow.. and hopefully will see this thread ;)

1) Yeap. I think i asked quite a dumb question if RIPv1 support route summary when it didn't even support sending of netmask in the 1st place... I am sorry, sometimes i think studying too much dull my brain cells to the extend i can't think at basic level.. Apologies for that. ;[

2) What are host route ?

The ip summary-address rip routerconfiguration command causes the router to summarize a given set of routes learned via RIP Version 2 or redistributed into RIP Version 2. Host routes are especially applicable for summarization.

I googled and the only conclusion i can draw is /32 network in the routing table are host route . Initially i am thinking , it is a route that connect to 1 host only. But if it is /32, does it mean it is connected to itself (since the 1 and only availalbe ip is going to be assigned to the interface) ?   Why ?

3)

The only change in the summary route would be its metric. As Bilal explained, in Cisco implementation, a summary address inherits the lowest metric from among the routes that were summarized. If this minimal metric changed, so would the metric of the summary.

For RIPV2, the metric is hop count.
Does that mean if i send out a summary route through an interface to the adjacent router, the metric is going to be the lowest hop count among all the routes that were summarized ?. I have thought the metric is always going to be 1 permanently.   From the advertising router (sending the summary route) to the receiving router.

e.g if i am doing to send a summary route for the 192.168.0.0/22 network (in bold below) out of
10.10.30.0 network at fa0/1, the metric of summary route will be 1 (lowest) ?

     10.0.0.0/24 is subnetted, 3 subnets
C       10.10.10.0 is directly connected, FastEthernet0/0
C       10.10.20.0 is directly connected, Ethernet0/1/0
C       10.10.30.0 is directly connected, FastEthernet0/1
R    192.168.0.0/24 [120/2] via 10.10.10.1, 00:00:21, FastEthernet0/0
R    192.168.1.0/24 [120/2] via 10.10.20.1, 00:00:27, Ethernet0/1/0
R    192.168.2.0/24 [120/1] via 10.10.10.1, 00:00:21, FastEthernet0/0
R    192.168.3.0/24 [120/1] via 10.10.20.1, 00:00:27, Ethernet0/1/0

is the sending router going to add 1 to the metric (1+1 =2) before sending out ?

4) Well, this is a limitation in Cisco's implementation, not in the design of RIPv2 itself.

Peter - do you mean RIPv2 itself / in other non-cisco routers, allow supernetting in route summarization ?

Regards,
Noob

[SJ K]

Hi all,

Pushing this up so that Peter and the rest can see my latest questions ;)

Regards,
Noob

[Peter Paluch]

Hi,

1) Yeap. I think i asked quite a dumb question if RIPv1 support route summary when it didn't even support sending of netmask in the 1st place... I am sorry, sometimes i think studying too much dull my brain cells to the extend i can't think at basic level.. Apologies for that. ;[

No need to apologize. In fact, I had to stop and think for a while about the way RIPv1 and manual summarization could coexist before I concluded they are mutually exclusive. Now it all makes sense but I actually haven't thought of that in quite a while.

2) What are host route ?

Host routes are routing table entries whose netmask is /32 for IPv4 and /128 for IPv6. Effectively, a host route describes a single particular IP address and a path toward it. You can see host routes in several occassions:

• Loopback interfaces are commonly configured as single-addressed interfaces with a /32 netmask.
• In 15.x IOSes, each router's own IP address is shown in the routing table as an 'L' (local) route with a /32 netmask to make sure that packets for the router itself are not routed elsewhere.
• On links that use the Point-to-Point Protocol (PPP), the neighbor's address is installed as a host route into the routing table automatically.
• When running OSPF using a so-called point-to-multipoint network type, instead of the common network each router is connected to, it advertises the router's own IP address on that network using a /32 netmask.

Host routes can either refer to local interfaces (in the case of loopbacks, for example), or they can refer to remote hosts (for example a /32 loopback of another router learned via a routing protocol). There is in fact nothing special with host routes, they're just like any other route - either directly connected or remote, learned via a routing protocol or statically configured. Note, however, that a router usually allows only a loopback interface to be configured with a /32 netmask. Other interfaces need to use netmasks /31 or shorter because they in fact allow the router to connect to at least one more device which needs an address for itself.

e.g if i am doing to send a summary route for the 192.168.0.0/22 network (in bold below) out of 10.10.30.0 network at fa0/1, the metric of summary route will be 1 (lowest) ?

Yes, the summary route's initial metric will be 1 but when the router advertises this summary, it will add 1 to the initial metric just like with any advertised RIP routes. So the neighbor would learn the summary with the metric of 2.

Peter - do you mean RIPv2 itself / in other non-cisco routers, allow supernetting in route summarization ?

I am not entirely sure about other vendors but let put it this way: RIPv2 as a protocol (messages, their format, processing the information) has absolutely no limitations on this. It is only about what the operating system on a router allows you to put into those messages. On Cisco routers, IOS does not allows supernetting when doing route summarization. However, as I indicated earlier, you can actually force a router to send a supernet using RIPv2 if you define it as a static route and redistribute it into RIPv2.

Best regards,
Peter

[SJ K]

Hi Peter,

Thanks for replying to this thread.

Host routes are routing table entries whose netmask is /32 for IPv4 and /128 for IPv6. Effectively, a host route describes a single particular IP address and a path toward it. You can see host routes in several occassions:

Loopback interfaces are commonly configured as single-addressed interfaces with a /32 netmask.

[Noted. Been seeing that as well]

In 15.x IOSes, each router's own IP address is shown in the routing table as an 'L' (local) route with a /32 netmask to make sure that packets for the router itself are not routed elsewhere.

[Alan] By meaning of each router's own IP address -> do you mean each router's interface's IP address ?  From my knowledge now , the only use to reach a router's interface's IP is to access the management (telnet, ssh in). Is there any other scenarios whereby the router's interface's IP need to be reached and it must be "taken care" of, to be make sure that the packets are not routed elsewhere; or can rather, how can it be routed elsewhere ?

On links that use the Point-to-Point Protocol (PPP), the neighbor's address is installed as a host route into the routing table automatically.

[Alan] I am actually confused about this. As per what you have mentioned later -> "Other interfaces need to use netmasks /31 or shorter because they in fact allow the router to connect to at least one more device which needs an address for itself."  

For any connection with an adjacent router or host ,  at least 2 IPs from the same subnet must be used, 1 from the current router interface, and 1 for the adjacent router's interface or adjacent host.    I have not studied on PPP yet, does PPP connection require the 2 connecting point to be in the same subnet ?  What will be the route type for this host route ?

When running OSPF using a so-called point-to-multipoint network type, instead of the common network each router is connected to, it advertises the router's own IP address on that network using a /32 netmask.

[Alan] Not very sure of how point to multipoint work but I do have some questions on the type of network nbma etc.. Will post them in another thread.

Host routes can either refer to local interfaces (in the case of loopbacks, for example), or they can refer to remote hosts (for example a /32 loopback of another router learned via a routing protocol). There is in fact nothing special with host routes, they're just like any other route - either directly connected or remote, learned via a routing protocol or statically configured. Note, however, that a router usually allows only a loopback interface to be configured with a /32 netmask. Other interfaces need to use netmasks /31 or shorter because they in fact allow the router to connect to at least one more device which needs an address for itself.

[Alan] It seems to be that, a /32 host route in routing, can only be

i) Local interface,

ii) Local Loopback interface

iii) Remote host's loopback interface learn through routing advertisement.

iv) a static route to just a single remote host 

iiv) is there any /32 route that is non of the above nature ? (e.g. a direct connection to a host -- don't think so.. need at least /31)

Yes, the summary route's initial metric will be 1 but when the router advertises this summary, it will add 1 to the initial metric just like with any advertised RIP routes. So the neighbor would learn the summary with the metric of 2

So if the lowest metric is 2 hops among the routes to be summarized, the router will add 1 to its metric, and send out 3 to the neighbor.  Right ?
In another word as well, summarized route hop metric may not be accurate (due to the fact it will use the lowest metric as the current hop count)

Regards,
Alan

- See more at: https://supportforums.cisco.com/discussion/12519381/ip-summary-route-rip-and-eigrp#sthash.F4Po6O32.dpuf

[Jon Marshall]

Bilal

Maybe the super experts like Peter and Jon can explain better and correct me if im wrong :)

You obviously wrote this before the PBR question then :-)

Jon

[Bilal Nawaz]

Peter, in my eyes, its no exaggeration, most of your posts, I have bookmarked for reference, and often find myself referencing back quite a lot, so great credit to you for that.

Hehe, Jon I have those moments probably more than you, and it's true, that I learn and take the nutritional information in your posts too.

[Jon Marshall]

Bilal

I think I have more of those moments but only because I post more answers :-)

And I suspect I have learned just as much from your posts as you have from mine.

Peter really is an expert (he is a professor in networking after all !) and I definitely agree with your comments about his posts.

I don't bookmark his posts but only because I like to live dangerously ie. I post into threads knowing full well if he is around he will most likely need to correct the mistakes I make :-)

Jon