Routing Loop - SVIs - eigrp, ospf
I created the following topology in PacketTracer as I work with similar configurations.
TOPOLOGY: http://img819.imageshack.us/img819/8838/vlan1.png
I wanted to test many different options and scenarios.
As you can see I have two L3 switches with some vlans, all routers and L3 switches are running EIGRP. Routers to L3 switches are connected using a routed port. Life is great, but when I disconnect the primary link (which in my situation is the serial link between RQ01 and RLON01, eigrp creates a loop. Why? I guess because I advertised all svi's on both switches. A few show commands to make it easier to understand (after removing the primary link)
SQ01# sh ip route
Gateway of last resort is 210.2.1.205 to network 0.0.0.0
1.0.0.0/30 is subnetted, 1 subnets
C 1.1.1.0 is directly connected, FastEthernet0/2
177.77.0.0/24 is subnetted, 1 subnets
D 177.77.77.0 [90/256005120] via 1.1.1.2, 00:08:30, FastEthernet0/2
210.2.1.0/24 is variably subnetted, 7 subnets, 4 masks
C 210.2.1.0/26 is directly connected, Vlan11
C 210.2.1.64/27 is directly connected, Vlan10
D 210.2.1.96/28 [90/205347840] via 210.2.1.131, 00:08:38, Vlan20
[90/205347840] via 1.1.1.2, 00:08:38, FastEthernet0/2
C 210.2.1.112/28 is directly connected, Vlan12
C 210.2.1.128/28 is directly connected, Vlan20
<code>
SQ01#sh ip eigrp neighbors
IP-EIGRP neighbors for process 1
H Address Interface Hold Uptime SRTT RTO Q Seq
(sec) (ms) Cnt Num
0 210.2.1.205 Fa0/24 11 00:02:26 40 1000 0 69
1 1.1.1.2 Fa0/2 13 00:02:26 40 1000 0 99
2 210.2.1.131 Vlan 14 00:01:54 40 1000 0 103
3 210.2.1.66 Vlan 13 00:01:54 40 1000 0 100
4 210.2.1.3 Vlan 12 00:01:53 40 1000 0 101
5 210.2.1.147 Vlan 10 00:01:53 40 1000 0 102
6 210.2.1.115 Vlan 12 00:01:53 40 1000 0 97
</code>
Traceroute from PC0 (bounces between two L3 switches)
<code>
PC>tracert 210.2.1.97
Tracing route to 210.2.1.97 over a maximum of 30 hops:
1 6 ms 8 ms 7 ms 210.2.1.1
2 15 ms 11 ms 13 ms 210.2.1.3
3 17 ms 18 ms 13 ms 210.2.1.1
4 15 ms 20 ms 22 ms 210.2.1.3
5 26 ms 18 ms 18 ms 210.2.1.1
6 28 ms 25 ms 31 ms 210.2.1.3
7 28 ms 29 ms 18 ms 210.2.1.1
8 37 ms 43 ms 22 ms 210.2.1.3
9 23 ms 35 ms 23 ms 210.2.1.1
10 41 ms 46 ms 29 ms 210.2.1.3
11 39 ms 43 ms 38 ms 210.2.1.1
12 37 ms 28 ms 51 ms 210.2.1.3
13 40 ms 50 ms 35 ms 210.2.1.1
14 44 ms 62 ms 47 ms 210.2.1.3
15 54 ms 55 ms 38 ms 210.2.1.1
16 50 ms 63 ms 63 ms 210.2.1.3
17 53 ms 70 ms 71 ms 210.2.1.1
18 75 ms 72 ms 74 ms 210.2.1.3
19 42 ms 45 ms 74 ms 210.2.1.1
20 63 ms 59 ms 63 ms 210.2.1.3
21 79 ms 58 ms 53 ms 210.2.1.1
22 84 ms 56 ms 87 ms 210.2.1.3
23 74 ms 76 ms 49 ms 210.2.1.1
24 82 ms 92 ms 63 ms 210.2.1.3
25 85 ms 86 ms 86 ms 210.2.1.1
26 93 ms 100 ms 62 ms 210.2.1.3
27 99 ms 99 ms 101 ms 210.2.1.1
28 69 ms 65 ms 102 ms 210.2.1.3
29 103 ms 81 ms 110 ms 210.2.1.1
30 115 ms 96 ms 119 ms 210.2.1.3
</code>
I tried to convert it to ospf and it was better, ospf did not create a loop but in the routing table I saw a few routes (on SQ01):<code>
O 177.77.77.0 [110/10001] via 210.2.1.66, 00:00:28, Vlan10
[110/10001] via 210.2.1.115, 00:00:28, Vlan12
[110/10001] via 210.2.1.131, 00:00:28, Vlan20
[110/10001] via 1.1.1.2, 00:00:28, FastEthernet0/2
</code>
Metric 10001 because I forced the second link to be 10kbs :)
I want to ask what the best practise is to avoid situations like this and why eigrp did not manage to figure it out ;-) I found two solutions:
1. Convert a routed port between two L3 switches to a trunk
2. Passive interface on all vlans
From the CCIE R&S Certification Guide Fourth Edition, page 198, Figure 6-4 Physical Topology Example Using MLS: The picture depicts having a dedicated routed port between the two layer 3 switches, as well as a dedicated trunk port. Technically you could do this with a dedicated vlan and vlan interfaces over the trunk.
Regarding the passive interfaces, I'd recommend passive-interface default and then no passive-interface on the specific routed ports. Unless there is a reason for sending EIGRP hellos to a client vlan, if there's no router there, why try to establish adjacency.
If you have a routed port instead of a layer 2 trunk with VLAN SVIs, the routed port can lead to slightly faster convergence times in the event of a link flap or port failure. This is due to the lack of spanning tree being involved in a routed interface. Definitely a plus.
With regard to EIGRP vs OSPF, there are pros and cons for each. EIGRP is an Advanced Distance vector, but borrows some features from link state protocols. An advantage to EIGRP is the unequalcost load balancing. However, OSPF offers greater scalability features with multiple areas and summarization at the ABRs.
The issue you are seeing with EIGRP in PacketTracer may be specific to PacketTracer. I've seen some weird glitches with the application. Hard to say without labbing it up with real gear...
That is exactly what I see, eigrp creates a loop when I keep disconnection the primary link...
Thanks Adam!
Thanks Adam. I think it is the best way, default passive-interface and use one vlan (I saw some recommendation that we should use our management vlan to do that) or routed ports to establish the neighbors. Any special reason why I should have both at the same time? (routed port and trunk) - one for routing and one for LAN I guess?
Am I right to say, that ospf was better than eigrp because eigrp is not a true link-state routing protocol and ospf had a map of the whole network
I was still surprised that eigrp could not figure it out ;-)