eigrp unequal what?!

silaghi · ‎03-25-2004

greetings... have a bit of an issue seeing any kind of load sharing with eigrp

tampa03r01#shr 10.40.2.5

Routing entry for 10.40.2.0/24

Known via "eigrp 100", distance 90, metric 2181120, type internal

Redistributing via eigrp 100

Last update from 172.25.10.46 on Multilink1, 08:55:40 ago

Routing Descriptor Blocks:

* 172.25.10.6, from 172.25.10.6, 08:55:40 ago, via Serial1/0:0.30

Route metric is 2181120, traffic share count is 48

Total delay is 20100 microseconds, minimum bandwidth is 1536 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 1

172.25.10.46, from 172.25.10.46, 08:55:40 ago, via Multilink1

Route metric is 4562432, traffic share count is 23

Total delay is 100100 microseconds, minimum bandwidth is 1280 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 1

i was told by tac that seeing traffic share count 48/23 on these routes meant that out of 71 packets, 48 would go to serial and 23 to multilink

started with a very basic variance 2 config and running this now

router eigrp 100

variance 3

redistribute static

network 10.0.0.0

network 172.25.0.0

maximum-paths 2

no auto-summary

---------

far side capture and config:

newyork06r01#shr 10.10.4.7

Routing entry for 10.10.4.0/24

Known via "eigrp 100", distance 90, metric 2181376, type internal

Redistributing via eigrp 100

Last update from 172.25.10.45 on Multilink1, 08:56:38 ago

Routing Descriptor Blocks:

* 172.25.10.5, from 172.25.10.5, 08:56:38 ago, via Serial1/0:0.30

Route metric is 2181376, traffic share count is 48

Total delay is 20110 microseconds, minimum bandwidth is 1536 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 2

172.25.10.45, from 172.25.10.45, 08:56:38 ago, via Multilink1

Route metric is 4562688, traffic share count is 23

Total delay is 100110 microseconds, minimum bandwidth is 1280 Kbit

Reliability 255/255, minimum MTU 1500 bytes

Loading 1/255, Hops 2

router eigrp 100

variance 3

redistribute static

network 10.0.0.0

network 172.25.0.0

maximum-paths 2

no auto-summary

... tried various attempts at manipulating metrics, even tried adjusting k2 value (VERY briefly).. have had equal delay/bw statements, now at true representation... point is, no matter what happens cannot get routers to utilize multilink1 interface when generating traffic, on either side.. ..suspected ip cef as culprit so disabled it to no avail.. presently with this config have about 800k bothways in usage on serial and like 43bps on multilink... also tried changing from mlppp encap to hdlc with no change.

any suggestions or previous experience with this problem greatly appreciated... running 12.2(15)T11 at tac's recommendation down from 12.3 as unknown bug was cited as their explanation for these symptoms

dathaide · ‎03-26-2004

hi

the variance command and unequal load balance will work if your backup route Admin distance is less than your current route feasible distance.

thanks

silaghi · ‎03-26-2004

thanks.. if i understand you correctly you are saying that load sharing will not work if the backup route (route without the * in the show ip route output) has a admin distance (metric shown in the route table) that is a smaller number numerically than what shows up as the FD number/metric in the show ip topology for that same route... ?

what are your thoughts on this:

when i set both interfaces to exact same values for bandwidth statement and delay, only one route is selected for all traffic despite the fact that show ip route displays both interfaces with the exact same metric... the only interface that will be used at all times for that route is the one with the * next to it, provided it is the first route displayed on the output... right now i have 700k of traffic and the exact same metrics on both interfaces and all 700k is taking one path, as more traffic is generated, that same interface increases from 700k to 800k with nothing happening on the other interface possessing the exact same metric

bjellig · ‎03-26-2004

The decision on which route (among equal-cost routes) also relies on your packet switching configuration (process/fast-switching/CEF).

This link describes these details of load-balancing route selection process.

http://www.cisco.com/en/US/tech/tk365/tk80/technologies_tech_note09186a0080094820.shtml

silaghi · ‎03-26-2004

thanks again for your reply.. right... thing is, i have disabled cef globally, prior to that adjusted from per packet to per destination cef config at interface level.. is there anything more to do re: cef to take it out of equation, other than disabling it globally?

dathaide · ‎03-26-2004

hi

Yes the Reported distance of the backup route should be less than the Feasible distance for the primary route for it to be a successor

do a search on cisco website for unequal load balancing with variance and they will give u an example. Do a Sho ip route eigrp topology and you should see the feasbile successor for a route. If you don't then unequal load balancing is not working correctly. Also are you doing per packet or per destination. Per destination is enabled by default unless you put in the no ip route cache statement.

Sorry i don;t have the link handy

thanks