BGP Sessions with HKIX Found Flapping from 22:10 Dec 17, 2013 and 02:00 Dec 18, 2013
The BGP sessions between HKIX and several HKIX participants were found flapping between 22:10 Dec 17, 2013 and 02:00 Dec 18, 2013.
After investigation, we discovered a HKIX participant conducted its router interface re-configuration last night without proxy ARP DISABLED causing some peering IP addresses binding to the same MAC address at the Route Servers’ ARP tables; which then caused the flapping.
The participant’s port was then shut down and the wrong ARP caches of both Route Servers were cleared; the BGP sessions returned normal at around 02:00 on Dec 18, 2013.
The participant’s port is currently put in a testing VLAN for joint investigation and we will only put its port back to the production VLAN when the problem/issue is completely solved.
I would also like to take this opportunity to remind all HKIX participants to ensure that proxy ARP is DISABLED on your router interface towards HKIX at all times.
Thanks for your attention.