cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Announcements
504
Views
20
Helpful
5
Replies

Do customers CCU dropped because BGP hold time?

Hi,

I'm an ISP, when one BGP peer went down, I found my customers ConCurrent User dropped. I would like to ensure is it root cause came from BGP hold time expire, right?

Thank you very much.

5 REPLIES 5
Giuseppe Larosa
Hall of Fame Master

Hello @ปลาวาฬทราย RMUTT CPE IX ,

what do you mean with ConCurrent Use   

do you mean user connections ?

Any form of collaboration clients ? like webex ?

 

In any case when the BGP peer goes down all the cstomer's related prefixes are withdrawn with impact on user traffic flows whatever they are.

if the customer is multi homed to another ISP it can try to re-route via other ISP but only if the customer owns its own public address blocks.

In case of mulihoming reaction starts after receiving BGP notification that the session has been turned down from the other side for the expiration of the BGP hold time.

 

Hope to help

Giuseppe

 

 

When eBGP peer link 1 or 2 down, does the game servers CCU dropped because BGP hold time?

 

CCU.jpg

Thank you very much.

Hello,

 

it is unlikely and sort of very outdated to wait for the BGP hold timer to expire. Cisco's BGP 'fast-external-fallover' feature (enabled by default) terminates external BGP sessions without waiting for the hold-down timer to expire when the peer goes down. If not that feature, what do you (as the ISP) or the customer have configured as failover mechanism ?

 

Configuration here:

 

Spoiler
   template peer I-Peer
    description Neighbor to Route Server IX IPV4
    remove-private-as
    address-family ipv4 unicast
      send-community both
      route-map I-IN in
      route-map I-OUT out
      maximum-prefix 15000 warning-only
      soft-reconfiguration inbound always
!
  neighbor x.19.61.1 remote-as x652
    inherit peer I-Peer
    no shutdown
    address-family ipv4 unicast

 

Thank you very much.

Hello,

 

as stated, the 'fast-failover-external' feature is enabled (by default), so the peering relationship is reset immediately, without waiting for the hold-down timer to expire. What protocol(s) are the game servers using ? There might not be any tolerance for packet loss or delay at all, so that is why they could have been reset; even when the BGP failover was immediate, there is always a delay, albeit very small.