[c-nsp] Strange behavior in ECMP of OSPF
Matt Addison
maddison at iquest.net
Sun Aug 21 09:12:04 EDT 2005
Don't know if it'll help or not, but to prevent quagga becoming DR in your OSPF areas:
ip ospf priority 0
On your quagga boxes, setting priority to 0 will prevent them from ever becoming the DR, so DR/BDR roles should fall to your cats.
http://www.quagga.net/docs/docs-info.php#OSPF-interface
-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Joe Shen
Sent: Sunday, August 21, 2005 5:15 AM
To: cisco-nsp at puck.nether.net
Subject: [c-nsp] Strange behavior in ECMP of OSPF
Hi,
I met a problem which conflict with my understanding of ECMP in OSPF. Two OSPF boder router form different route table on six equal cost path to the same IP address!
The following is my network configuration
Server6
\ (trunk carring Vlan3)
Cata6509-a ------------------- Cata6509-b
| \ /
Server5 \ /
\ /
\-------------------------/
| Cata4506 |
-------------------------
/ | | \
/ | | \
Server-1 Server-2 Server-3 Server-4
All Servers running OSPF daemon which announce address
10.101.172.35 beside its interface address (10.101.172.128/28). All servers are put into Vlan 3, while Catalyst6509-a and Catalyst6509-b have Vlan3 interface ( IP address is 10.101.172.129, 10.101.172.130). Catalyst6509-a, catalyst6509-b and the six servers form OSPF NSSA area 10.101.172.128, Catalyst6509-a & catalyst6509-b are border routers which belong to area0 too.
Both Catalyst6509 is configured as:
=============================================
router ospf 18181
router-id 10.71.101.1
log-adjacency-changes
auto-cost reference-bandwidth 10000
area 10.101.172.128 authentication message-digest area 10.101.172.128 nssa default-information-originate no-summary redistribute connected metric-type 1 subnets redistribute static metric-type 1 subnets network 10.101.172.128 0.0.0.15 area 10.101.172.128 maximum-paths 8 distribute-list 18 in Vlan3
6509-a#sh ip access-lists 18
Standard IP access list 18
permit 10.101.172.35 (626 matches)
permit 10.101.172.47 (626 matches) check=160
permit 10.101.172.128, wildcard bits 0.0.0.31 (58
matches) check=21160
deny any (21160 matches)
6509-a#
================================================
And, Vlan 3 on both Catalyst6509 is configure like:
========================================
6509-a#sh run int vlan 3
Building configuration...
Current configuration : 291 bytes
!
interface Vlan3
description Server-Cluster-1
ip address 10.101.172.130 255.255.255.240 no ip redirects no ip unreachables ip route-cache flow ip ospf message-digest-key 19 md5 7 02070A4208071C35 ip ospf priority 10 end =====================================================
After setting up 6509-a , 6509-b and the servers, I noticed all of them establish ospf neighbor relationship successfully, but DR is not on 6509 but on one of the servers! On the other hand, there are six equal cost routing path on Catalyst6509-a to address 10.101.172.35, but there is ONLY ONE route to that destination on Catalyst6509-b!!!
================================================
6509-a#sh ip route 10.101.172.35
Routing entry for 10.101.172.35/32
Known via "ospf 18181", distance 110, metric 20, type NSSA extern 2, forward metric 10
Last update from 10.101.172.137 on Vlan3, 00:42:24 ago
Routing Descriptor Blocks:
10.101.172.133, from 10.101.172.133, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
10.101.172.134, from 10.101.172.134, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
10.101.172.135, from 10.101.172.135, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
10.101.172.136, from 10.101.172.136, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
10.101.172.137, from 10.101.172.137, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
* 10.101.172.138, from 10.101.172.138, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1
6509-a#
6509-b#sh ip route 10.101.172.35
Routing entry for 10.101.172.35/32
Known via "ospf 100", distance 110, metric 20, type NSSA extern 2, forward metric 10
Last update from 10.101.172.136 on Vlan3, 00:42:24 ago
Routing Descriptor Blocks:
10.101.172.136, from 10.101.172.136, 00:42:24 ago, via Vlan3
Route metric is 20, traffic share count is 1 6509-b# ==================================================
If I withdraw 10.101.172.35 from 10.101.172.136, route on 6509-b will move to another server, never will it see all the servers at one time!
Browsing ospf database on 6509-b, it shows the database stores all LSA announcing 10.101.172.35 from the six servers.
Why? Is there any configuration problem?
I run quagga0.98.3 on those servers, the 6509 IOS version is:
============================================
6509-a:
6509-a# sh ver
Cisco Internetwork Operating System Software IOS (tm) c6sup2_rp Software (c6sup2_rp-PSV-M), Version 12.1(13)E15, RELEASE SOFTWARE (fc1) Technical Support: http://www.cisco.com/techsupport Copyright (c) 1986-2004 by cisco Systems, Inc.
Compiled Wed 11-Aug-04 05:15 by pwade
Image text-base: 0x40008C00, data-base: 0x417A8000
ROM: System Bootstrap, Version 12.1(11r)E1, RELEASE SOFTWARE (fc1)
BOOTLDR: c6sup2_rp Software (c6sup2_rp-PSV-M), Version 12.1(13)E15, RELEASE SOFTWARE (fc1)
6509-b#sh ver
Cisco Internetwork Operating System Software IOS (tm) MSFC2 Software (C6MSFC2-PSV-M), Version 12.1(13)E15, RELEASE SOFTWARE (fc1) Technical Support: http://www.cisco.com/techsupport Copyright (c) 1986-2004 by cisco Systems, Inc.
Compiled Wed 11-Aug-04 09:27 by pwade
Image text-base: 0x40008C00, data-base: 0x41884000
ROM: System Bootstrap, Version 12.1(4r)E, RELEASE SOFTWARE (fc1)
BOOTLDR: MSFC2 Software (C6MSFC2-PSV-M), Version 12.1(13)E15, RELEASE SOFTWARE (fc1)
======================================
Each word will be highly appreciated!
Joe
___________________________________________________________
雅虎免费G邮箱－中国第一绝无垃圾邮件骚扰超大邮箱
http://cn.mail.yahoo.com
_______________________________________________
cisco-nsp mailing list cisco-nsp at puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
More information about the cisco-nsp
mailing list