I have 2 eBGP peers over a 64k leased line. One of the peers, say router A
periodically sends a BGP notification of "hold time expired" to the other
peer router B and it appears to happen quite randomly.
Whenever this happened, on router B i could actually see a syslog message
"BGP-3-NOTIFICATION: received from neighbor <router A's IP> 4/0 (hold time
expired) 0 bytes"
Does this indicate that the link is actually not down (since router B
received the BGP notification from A)? How then, could router A have missed
3 consecutive keepalives that router B sent?
I've checked the timers, and the peers negotiated the keepalive to be 10
seconds, holdown time to be 30 seconds; nothing out of the ordinary here.
Hmm . .. good point there.
My debug shows that there's not much else happening after the BGP session
establishment. There're only a few routes exchanged after BGP session open.
It's only keepalives and some minor BGP housekeeping traffic thereafter.
Could it be that the actual application data delaying BGP keepalives and
when tcp is timed out consistently, the peer thinks that it's dead? I might
have to look into QOS to prioritize the BGP traffic.
"www.networking-forum.com" <email@example.com> wrote in message
> 64k? Could it be that your BGP updates are swamping that line?