ping causing bgp peering over trunk links to fail

Hey Guys, I got a real peculiar situation.

​

We have a EoL Catalyst 6500 (with sup720) 3 links:

​

A. A layer3 link that forms a bgp peering btw the 6500 and a 6800 (2x 10G port channel)

B. A trunk link with multiple bgp peerings btw the 6500 and a N7K (1x 10G link)

C. Another trunk link with multiple iBGP peering between the 6500 and another 6500

​

​

PROBLEM:

A few days ago, with no config changes or anything on the 6500. all of a suddent, all the bgp peerings over the 2 trunk link (B & C) went to "idle/active" state.

while the bgp peering over the layer 3 link is fine.

my SVIs on my 6500 are up and also the neighbors (over the trunk links) are reachable (via ping). but why does bgp peering just fail?

​

this is what bgp summ looks like:

​

Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd

x.x.x.x 4 200 11 2820 509167 0 0 00:05:37 3 (over link A)

y.y.y.y 4 300 78 39 509167 0 0 00:05:24 104 (over link B)

z.z.z.z 4 40 41 74904 509184 0 0 00:07:52 1 (over link C)

a.a.a.a 4 40 335 39 509184 0 0 00:08:23 1965 (over link C)

b.b.b.b 4 40 4909 71654 509184 0 0 00:08:24 15510 (over link C)

c.c.c.c 4 500 132693 40 509183 0 0 00:05:37 50220 (over link B)

​

​

this is what bgp summ looks like after the failure, only bgp on layer3 link is up.

​

Neighbor V AS MsgRcvd MsgSent TblVer InQ OutQ Up/Down State/PfxRcd

x.x.x.x 4 200 11 2820 509167 0 0 00:05:37 3 (over link A)

y.y.y.y 4 300 78 39 509167 0 0 00:05:24 ACTIVE (over link B)

z.z.z.z 4 40 41 74904 509184 0 0 00:07:52 ACTIVE (over link C)

a.a.a.a 4 40 335 39 509184 0 0 00:08:23 ACTIVE (over link C)

b.b.b.b 4 40 4909 71654 509184 0 0 00:08:24 ACTIVE (over link C)

c.c.c.c 4 500 132693 40 509183 0 0 00:05:37 ACTIVE (over link B)

​

​

​

TROUBLESHOOTING:

​

a reload of the catalyst fixed the bgp peerings as all bgp peerings over the trunk links come back up. but at random times, all bgp peerings over the trunk link went to active state again. and a reload would fix it AGAIN.

​

​

i found out that running ping to any remost host triggered the bgp peerings over the trunk links to crash and go to active. even if it was a ping that didnt transverse the trunk links.

​

i replaced the catalyst 6500 with a 6807XL (with dual sup2T) and the problem still exists. and as with the 6500, a reload would bring up all bgp peerings.

and again i noticed that any ping/icmp traffic that comes in from any interface crashed all bgp peerings over trunk links.

​

if no pings/icmp traffic come into the 6807xl, the bgp peerings would stay up, but once ping was run it immediately crashed the bgp peerings on the trunks

​

I have never seen anything like this before. 6807xl is running 155.1.SY4

this is weird. please has anyone experienced this issue before?

​

thanks!!