Strange behaviour

[cisco_lad2004]

Hi all

I have 2 core routers connected toegther via a 6509.

R1---6509---R2

on R2, I placed 2 loopbacks, one of them pings fine from R1, the other does not. routing parameter are exactely the same.

then I noticed on my 6509 the following error message, coudl this be teh explanation ? teh second(bad) Loopback represents a new range added to our network and my concern is that we have exhausted out cef tags.

Oct 25 06:33:09: %MLSCEF-SP-7-FIB_EXCEPTION: FIB TCAM exception, Some entries will be software switched

so in this case, I dont even dare to reload as the bad IP range might start working but other good ones will stop depending on which comes 1st.

The othert explanation coudl be CPU spikes...but why only for the badrange ?

Any thoughts would be great help!

TIA

Sam

[lgijssel]

First of all I would carefully inspect the ip routing tables to make sure that this is not a duplicate range.

If it is CEF-related as you suggest, you could try to remove the first loopback range (temporarily) and check if the second one becomes accessible then.

Regards,

Leo

[cisco_lad2004]

Hi Leo

Thanks for te reply.

good point about duplicate routing or assigned ip. but then I shoudl expect to see 50% drop excately which is not the case.

with reagrds to Lp removal, sadly all is on production.

Cheers

Sam

[Richard Burts]

Sam

The error message indicates that some packets would be software switched. This would impact CPU load but would not prevent connectivity. So I think that you need to look for a different problem.

Your first posting just said that the second loopback did not ping fine. Can you be more specific about the problem? Can you ping it at all? Can you traceroute to it? Is there perhaps an access list involved which impacts it?

Perhaps you could post the configuration of both loopbacks and the configuration of your routing, and the output of show ip route on the second router? If we had more information to work with perhaps we could find the problem.

HTH

Rick

[cisco_lad2004]

Hi Rick

for bad loopback, I loose packets, about 20%.

this behaviour is consistent.

-no error on transmission

-no Q drops.

-no congetsion or policy-map drops.

-no ACL's

-same routing, same outgoing intetfaces.

6509 advertise a default route to R1 and 6509 has BGP session to R2. this is all a production network where 6509 is a bgp peering box and R2 a transit provider.

I agree, CEF resources might be a red herring.

Thanks

Sam

[bella]

could you pls provide output of the sh ip route your_bad_loopback_address_here command from R1?

[cisco_lad2004]

Hi Bella

From R1, both Lps are pointing to default route ( 0.0.0.0/0 received from 6509)

on 6509:

SOISP01#sh ip route 85.80.194.197

Routing entry for 85.80.0.0 255.252.0.0

Known via "bgp 9158", distance 20, metric 5

Tag 6785, type external

Last update from 212.242.27.9 21:08:30 ago

Routing Descriptor Blocks:

* 212.242.27.9, from 212.242.27.9, 21:08:30 ago

Route metric is 5, traffic share count is 1

AS Hops 1

SOISP01#sh ip route 212.242.3.103

Routing entry for 212.242.0.0 255.255.0.0, supernet

Known via "bgp 9158", distance 20, metric 5

Tag 6785, type external

Last update from 212.242.27.9 21:08:43 ago

Routing Descriptor Blocks:

* 212.242.27.9, from 212.242.27.9, 21:08:43 ago

Route metric is 5, traffic share count is 1

AS Hops 1

SOISP01#

[cisco_lad2004]

Hi all

I have found some interesting info:

http://www.cisco.com/univercd/cc/td/doc/product/software/ios123/123cgcr/ipv6_r/ipv6_09g.htm#wp1983609

The new configurations are applied after a system reload only and do not take effect if a switchover occurs.

In RPR mode, if you change and save the maximum-routes configuration, the redundant supervisor engine reloads when it becomes active from either a switchover or a system reload. The reload occurs 5 minutes after the supervisor engine becomes active.

Use the show mls cef maximum-routes command to display the current maximum routes system configuration.

Examples

This example shows how to set the maximum number of routes that are allowed per protocol:

Router(config)# mls cef maximum-routes ip 100

This example shows how to return to the default setting for a specific protocol:

Router(config)# no mls cef maximum-routes ip

this will be tried tonight during maintenance window.

Cheers

Sam

[cisco_lad2004]

Hi all

I can confirm, my issue was fixed when mls cef max routes was increased.

what did I learn ? there is never a strange behvaiour and laways an unknown logical reason...we don't know about.

Thanks to all

Sam