routes deleted 15sec after the BGP hold time expired

[samachan]

Hi expert,

After bgp hold time expired, the router doesn't delete the routes from that bgp peer at once, but wait for 15 secs. What is the reason? can we get rid of the delay?

Thanks,

Samantha

[Giuseppe Larosa]

Hello Samantha,

Cisco implementation of BGP uses some timers and it is not totally event driven.

There is the BGP scan timer that is different for address family ipv4 or vpnv4. In other words BGP table is scanned every N seconds looking for changes.

see

http://www.cisco.com/en/US/docs/ios-xml/ios/iproute_bgp/command/bgp-a1.html#GUID-41436812-87CB-446C-9525-D12E84690DD6

bgp scan-time can change this

you may be also interested in BGP next-hop tracking if the context is iBGP

Hope to help

Giuseppe

[samachan]

Hello Giuseppe,

Thanks for your reply. I disabled the NHT and changed the scan timer, but it still takes 15 secs every time.

The session is EBGP. Is there any other possible reason?

Regards,

Samantha

[Giuseppe Larosa]

Hello Samantha,

NHT should provide benefits can you check by using

router bgp

address-family ipv4 unicast

bgp next-hop trigger enable

bgp next-hop trigger delay 2

How do you perform the checks on the routes?

Are you using debug ip routing + ACL that specifies the interesting routes ?

debug ip bgp events ?

I remember there are other timers involved, but I cannot find a reference in this moment.

I'm sorry

Hope to help

Giuseppe

[samachan]

Hi Giuseppe,

I changed the delay to 2 and captured “debug ip bgp event” in the attachment.

In the debug,

67.1.1.5 is the bgp peer over the main WAN interface

10.65.4.65 is the bgp peer over the IPSec SVTI

I shutdown the main WAN interface and the bgp over the IPsec SVTI expired. Will the “shutdown” action affect the behavior? I am simulating the setup and cannot use real equipment at the moment.

The time between the following two messages is always 14-15 sec. BGP deleted the routes when we saw %BGP_SESSION-5-ADJCHANGE.

May 9 16:22:42.363: %BGP-5-ADJCHANGE: neighbor 10.65.4.65 Down BGP Notification sent

May 9 16:22:56.847: %BGP_SESSION-5-ADJCHANGE: neighbor 10.65.4.65 IPv4 Unicast topology base removed from session BGP Notification sent

Please share your thoughts. Thanks.

[Giuseppe Larosa]

Hello Samantha,

BGP has its own BGP table that is different from the IP routing table.

You are interested in measuring the time it takes for the router to remove the routes learned on the failed eBGP session from the IP routing table.

The message that we see is likely related to removal of routes from BGP table ( a.k.a. RIB) and not from IP routing table.

Adding debug ip bgp update and debug ip routing with an ACL that specifies an IP prefix learned on the failed eBGP session could be helpful.

Hope to help

Giuseppe

[samachan]

Thanks for your advice. I am getting some real boxes and will test again.