Load balancing with serial and ethernet

[noran01]

We currently have the 3640 with a t1 connecting via serial interface and a handoff of a frac-t on an ethernet interface. I wanted to load balance this out using cef. I got the config running, but when will traffic switch onto the frac-t? i have per packet load balancing configured on both interfaces. but it seems the ethernet inerface is never utilized. can this even be done... i have two default router statements entered for each gateway.

[Harold Ritter]

you should be able to load-balance if you have the right configuration. Could you provide us with a "sh ip cef 0.0.0.0 int".

Thx

[noran01]

When i do that command this is the output:

0.0.0.0/32, version 0, receive

Here is the config that is under my e3/0 and s0/0 (the two interfaces i am trying to load balance)

interface Serial0/0

description :T-1 circuit

ip unnumbered FastEthernet0/0

ip load-sharing per-packet

interface Ethernet3/0

description :frac t-1

ip address x.x.x.x 255.255.255.240

ip load-sharing per-packet

full-duplex

Thanks...

[preddyi]

can you check for

show cef drop

show ip cef internal

sh ip cef 0.0.0.0 0.0.0.0 detail

[noran01]

Here is the output:

DR2-R1#sh cef drop

CEF Drop Statistics

Slot Encap_fail Unresolved Unsupported No_route No_adj ChkSum_Err

RP 290731 1 0 52 0 0

DR2-R1#sh ip cef in

IP CEF with switching (Table Version 59), flags=0x0

28 routes, 0 reresolve, 0 unresolved (0 old, 0 new)

28 leaves, 15 nodes, 19132 bytes, 95 inserts, 67 invalidations

1 load sharing elements, 328 bytes, 1 references

universal per-destination load sharing algorithm, id DAD809D8

3 CEF resets, 1 revisions of existing leaves

refcounts: 4136 leaf, 4096 node

Adjacency Table has 7 adjacencies

0.0.0.0/0, version 30, attached, per-packet sharing

0 packets, 0 bytes

via Serial0/0, 0 dependencies

traffic share 1, current path

valid adjacency

via Ethernet3/0, 0 dependencies

traffic share 1

valid glean adjacency

0 packets, 0 bytes switched through the prefix

tmstats: external 0 packets, 0 bytes

internal 0 packets, 0 bytes

Load distribution: 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 (refcount 1)

DR2-R1#sh ip cef 0.0.0.0 0.0.0.0 detail

0.0.0.0/0, version 30, attached, per-packet sharing

0 packets, 0 bytes

via Serial0/0, 0 dependencies

traffic share 1, current path

valid adjacency

via Ethernet3/0, 0 dependencies

traffic share 1

valid glean adjacency

0 packets, 0 bytes switched through the prefix

[ruwhite]

The bandwidths on these two are probably completely different--how are you learning the default over each interface? Is it two statics, or learned through some protocol, or (?). Could you do a show ip route 0.0.0.0 and post it here?

Russ.W

[noran01]

I have two default routes set up (static)

ip route 0.0.0.0 0.0.0.0 s0/0

ip route 0.0.0.0 0.0.0.0 e3/0

[preddyi]

can you try using the next hop IP address insted just interface.

example:

ip route 0.0.0.0 0.0.0.0 192.168.0.1 s0/0

ip route 0.0.0.0 0.0.0.0 192.168.1.1 e3/0

[ruwhite]

I would definitely use the next hop instead of the interface, for several reasons. First, and foremost, pointing a default at a broadcast interface is never a good idea. It will result in arp cache entries for every possible ip address reached through the default route, and will probably mess up your load sharing, as well as cuase you numerous headaches in the area of memory consumption and processor utilization.

Use the next hops, as in the example above, and see what difference that makes.

Russ.W

[noran01]

ok I changed the default route for the ethernet interface to point to the next hop ip address instead and it seems some traffic is flowing over it, but not as much compared to the serial link...

Now when i do a sh cef drop i get this:

Router#sh cef drop

CEF Drop Statistics

Slot Encap_fail Unresolved Unsupported No_route No_adj ChkSum_Err

RP 1225 863 0 0 0 0

[noran01]

I see it working a little better now but its actually hurting performance....all our inet traffic is going through a pix which is doing PAT and its causing major slowdowns....

[subbarao.s]

Why dont you use Per-destination load balancing? You have 2 dissimilar bandwidth links, so u better use per destination. In CEF per destinaiton load blancing takes SOURCE+DESTINAITON hash, so you can have better usage and performance as well.

HTH