When BGP protocol is enabled on more than one Virtual Routers, it is possible that the new TCP connection request for BGP neighbor establishment could fail. There is one BGP process per Virtual router and each process creates one TCP listener socket for BGP port. Linux TCP implementation hashes the new connection to any of the listener socket, if there are more than one socket for the same TCP port. If the connection request is received on a BGP process that do not handle the Virtual Router corresponding to the new connection, then that would be dropped. BGP establishment time could become high with default connection retry timeout of 120 seconds. Reducing the connection retry timer to a lower value would reduce the peer establishment delay.
The event will occur when BGP protocol is added to a new Virtual Router. Increase the connection retry timeout value using the configure bgp neighbor <neighbor> connect-retry <value> command.
Notice
[VR %vrid%] A connect retry value of 30 seconds is recommended for BGP neighbors, when BGP neighbors are configured on more than one Virtual Router. The default of 120 seconds may result in longer than acceptable peer establishment times.
| Name | Type |
|---|---|
| vrid | VRID |