[f-nsp] Bigiron RX4 - IPV6 causing out of nexthop entries

Pieter Taks p.taks at nforce.com
Thu Dec 15 15:13:46 EST 2011


Hi Dunc,

Try: "maximum-paths 1"

This is just a follow-up regarding the ip nexthop table becoming full all
the sudden.

I was (unfortunately) able to replicate this problem:

1. I brought ~150 peering sessions online from public exchanges. All was
fine, no nexthop paths showing any sign of running into depletion.

2. While bringing up the last ~150 sessions the 8th path on the nexthop
table started to become full. With each ~3 sessions that got brought up
the table became full very fast, even before I was able to bring up all
150 of them. After bringing up the last 50 out of the total 300 sessions I
stopped. As the table would deplete very soon if I continued bringing up
the rest.

Paths  Total  Free  In-use
  1    2816   2515   301
  2    512    504    8
  4    512    512    0
  8    256    64     192

3. Ran this command on the router configure mode:
maximum-paths 1

Paths  Total  Free  In-use
1    2816   2498   318
  2    512    512    0
  4    512    512    0
  8    256    240    16


In the end the only working solution was to set the maximum-paths to 1.
After doing this the table cleared itself to normal state immediately.

I am not sure if this is 'the' solution we are looking for, but it beats
the crap out of having no solution.

--
Best regards,
Pieter Taks
NFOrce Entertainment BV




Op 15-12-11 19:15 schreef Dunc <dunc.lockwood at thebunker.net>:

>Hi Pieter & list
>
>Does anyone have anything to add to this thread at all yet? It looks
>like we might be running into a similar thing.
>
>I also thought it was DEFECT000323367, and after upgrading our code to
>2.8.0a, everything seemed fine, for approx. 1 month, but it has come
>back worse than ever now. We are also using IPv6.
>
>
>Cheers,
>
>Dunc
>
>
>On 13/08/11 15:24, Pieter Taks wrote:
>> Hi Wouter,
>> 
>> I did not do such thing. What would you suggest to cam-partition the
>> next-hop table like? As my concern is why it would need any change to a
>> higher value. As in general the in use of paths <=8 is approx 40. But
>> without any reason it jumps to 256 (or perhaps higher when adjusted with
>> cam-partitioning).
>> 
>> Paths Total Free In-use
>> 1 2816 2805 11
>> 2 512 512 0
>> 4 512 512 0
>> 8 256 0 256
>> 
>> 
>> 
>> To me it sounds like bug nr #RX 02.7.03 RN, which should be resolved in
>> 2.7.02k. See below.
>> 
>> 	
>> 	
>> 		Defect ID: DEFECT000323367
>> 			
>> 			
>> 			
>> 			
>> 			
>> 			Technical Severity: Medium
>> Summary: IP Cache-entry is not removed until either ARP entry is removed
>> or IP Cache entry is cleared.
>> Symptom: IPnext-hoptablemaybecomefullontherouter.
>> Probability: High
>> Feature: IPv4 Forwarding
>> 			Function: Next Hop Table
>> Reported In Release: RX 02.7.02
>> 			
>> 			
>> 			
>> 			
>> 			
>> 			Service Request ID: 265521
>> 
>> 
>> Thank you for your reply and thinking with me.
>> 
>> 
>
>
>-- 
>Dunc Lockwood
>Principal Network Engineer
>
>The Bunker Secure Hosting Ltd.
>Ash Radar Station
>Marshborough Road
>Sandwich
>Kent CT13 0PL
>
>T: 01304 814800
>F: 01304 814899
>E: dunc.lockwood at thebunker.net
>
>
>_______________________________________________
>foundry-nsp mailing list
>foundry-nsp at puck.nether.net
>http://puck.nether.net/mailman/listinfo/foundry-nsp





More information about the foundry-nsp mailing list