[c-nsp] BGP Update Timers

[ANSW (Anders Schwartzbach)]

Hi all

I am having a small testup, where I am playing with BGP update timers

I have one thing, which is driving me insane - 

CISCO2851 nr 1 Running BGP (local AS number 65412)
250 loopback interfaces
distributed into BGP
|
Cisco 7609 nr 1 Running BGP (local AS number 65500)
interface GigabitEthernet6/1
ip vrf forwarding noc

! The two 7609 is connected using ten gig
10.255.0.8      4 65500      85      97        1    0    0 01:14:35        0

Cisco 7609 nr 2 Running BGP (local AS number 65500)
interface GigabitEthernet6/1
ip vrf forwarding noc
!
CISCO2851 nr 2 Running BGP (local AS number 65413)

When I shutdown 10 interfaces on "CISCO2851 nr 1 Running BGP" I'll get the first update after very few seconds on "CISCO2851 nr 2 Running BGP" 

responder-ip# ("CISCO2851 nr 2 Running BGP")

Jul 12 08:38:21.057: RT: del 2.0.10.0/24 via 10.255.1.65, bgp metric [20/0]
Jul 12 08:38:21.057: RT: delete subnet route to 2.0.10.0/24
Jul 12 08:38:21.057: RT: NET-RED 2.0.10.0/24
Jul 12 08:38:52.109: RT: del 2.0.11.0/24 via 10.255.1.65, bgp metric [20/0]
Jul 12 08:38:52.109: RT: delete subnet route to 2.0.11.0/24
Jul 12 08:38:52.109: RT: NET-RED 2.0.11.0/24
Jul 12 08:38:52.109: RT: del 2.0.12.0/24 via 10.255.1.65, bgp metric [20/0]
Jul 12 08:38:52.109: RT: delete subnet route to 2.0.12.0/24
Jul 12 08:38:52.109: RT: NET-RED 2.0.12.0/24
Jul 12 08:38:52.109: RT: del 2.0.13.0/24 via 10.255.1.65, bgp metric [20/0]
....

What I don't get, is the delay between 08:38:21 and 08:38:52. For some reason I'll get the last 9 updates 30 secs after the first update.

In case I swap the sequence, and starts with 2.0.13.0 it is still the first interface I shoots down, which is updated first (30 sec. before the next).

The delay is found between "CISCO2851 nr 1 Running BGP (local AS number 65412)" and "Cisco 7609 nr 1 Running BGP (local AS number 65500)"
from 06:56:44.122 to 06:57:14.614.

rdkba1pco001#"Cisco 7609 nr 1 Running BGP (local AS number 65500)"
*Jul 12 06:56:44.122: RT(noc): del 2.0.103.0 via 10.255.1.58, bgp metric [20/0]
*Jul 12 06:56:44.122: RT(noc): delete subnet route to 2.0.103.0/24
*Jul 12 06:57:00.666: BGP-VPN(4): Label 58(table 16) free and chkpt OK
*Jul 12 06:57:14.614: RT(noc): del 2.0.100.0 via 10.255.1.58, bgp metric [20/0]
*Jul 12 06:57:14.614: RT(noc): delete subnet route to 2.0.100.0/24
*Jul 12 06:57:14.614: RT(noc): del 2.0.101.0 via 10.255.1.58, bgp metric [20/0]
*Jul 12 06:57:14.614: RT(noc): delete subnet route to 2.0.101.0/24
*Jul 12 06:57:14.614: RT(noc): del 2.0.102.0 via 10.255.1.58, bgp metric [20/0]
*Jul 12 06:57:14.614: RT(noc): delete subnet route to 2.0.102.0/24
....

The 2851 removes the routing instance at once. The update into the 7609 is the strange one.

Does anyone have an idea of which timer or function I am looking after - how come the first update is so fast?

In this scenario all timers has been set up to default (scan, keepalive etc.), in order to keep it simple.
 

Kind rgds Anders Schwartzbach