Traffic load balancing between multiple port channel

[KKS2287]

We are running 4 port-channel routed links between two tiers. These tiers are in two different ASs. We are load balancing the traffic between all the PC ECMP links. CEF table shows all the four links as the next hop. But when we looked at the link utilization graph for a period of week, we have found that only 2 links are being used and other 2 are unused on the outbound direction. Not sure what is causing this issue. We are using the default "ip cef load-sharing algorithm universal"

[Georg Pauwen]

Hello,

--> We are running 4 port-channel routed links between two tiers. These tiers are in two different ASs

What does that setup look like ? What exactly are the two tiers and two different ASs ? Post a schematic drawing of that topology.

[KKS2287]

Port-Chanel 1, 2, 3 and 4 are l3 links. We have eBGP running between the tiers. All the links are 10gigs. On T2 Spine 1, if i look at the cef for all southbound traffic, it shows all the four po's as the next hop. But for traffic outbound it uses only po1 and 4 Or po2 and 3. For some reason, po1 and 4 or po2 and 3 gets ignored.

t2_spine_1#show int port-channel 1 | i /sec
MTU 9100 bytes, BW 20000000 Kbit/sec, DLY 10 usec,
30 second input rate 105830000 bits/sec, 20383 packets/sec
30 second output rate 6000 bits/sec, 2 packets/sec
t2_spine_1#show int port-channel 2 | i /sec
MTU 9100 bytes, BW 20000000 Kbit/sec, DLY 10 usec,
30 second input rate 222767000 bits/sec, 41469 packets/sec
30 second output rate 18203000 bits/sec, 25458 packets/sec
t2_spine_1#show int port-channel 3 | i /sec
MTU 9100 bytes, BW 20000000 Kbit/sec, DLY 10 usec,
30 second input rate 233623000 bits/sec, 42580 packets/sec
30 second output rate 24861000 bits/sec, 25294 packets/sec
t2_spine_1#show int port-channel 4 | i /sec
MTU 9100 bytes, BW 20000000 Kbit/sec, DLY 10 usec,
30 second input rate 86561000 bits/sec, 17269 packets/sec
30 second output rate 3000 bits/sec, 1 packets/sec
t2_spine_1#

t2_spine_1#show ip cef 25.0.16.0
25.0.16.0/23
nexthop 10.64.0.1 Port-channel1
nexthop 10.64.0.5 Port-channel2
nexthop 10.64.0.9 Port-channel3
nexthop 10.64.0.13 Port-channel4

[Giuseppe Larosa]

Hello @KKS2287 ,

Are you running eBGP with the peers ?

You say:

>> We are running 4 port-channel routed links between two tiers. These tiers are in two different ASs. 

Have you configured for ebGP multipath , BGP always compare MED , BGP deterministic MED ?

What do your receive from the upstream providers only a default route or full tables or a default route and some prefixes ?

Are the two PC in use outbound going to the same AS ?

Hope to help

Giuseppe

[KKS2287]

Yes, we are running eBGP

[KKS2287]

I marked this discussion as solved by mistake. I have reached out to the support team to resolve this issue.

[ngkin2010]

Hi,

Grateful if you could share the root cause and solution. 

[KKS2287]

Sorry, I marked it has solved by mistake. I have not found the answer yet.