[c-nsp] GLBP failover

[szilard csordas]

hello,

I have an ordinary glbp configuration. R1 and R2 have an uplink to the
ISP (eBGP) and have an iBGP session between them.Behind them are 2x L3
switches.When the R2's BGP session was flapping I wanted to have all
traffic routed out from the site through R1, so I shut down the R2's
LAN interface. The virtual MAC of the active forwarder has been taken
over by the listener forwarder
(which became active after the shut down).

My question is: should not have the new forwared send out gratuitous
arp to update the swithces' arp table about the new virtual mac?

Because the glbp times out the MAC after 1h (configured), which has
been taken over from the faild forwarder. The L3 switches have the
default arp timeout (4h). So, after the lan interface shutdown for 1h
everything was fine.after that an outage occured. (Clearing the arp
table of the L3 switches solved the problem -  it wasn't easy to find
a guy who had an access to them).

from cisco:
"Any gratuitous ARP mechanisms on the GLBP group members for the
virtual IP address are disabled to eliminate conflict between members
(failover is an exception that will be addressed)."
Was not this a failover?


eBGP        eBGP
|                    |
|                    |
R1 -- iBGP -- R2
|                    |
|                    |
S1 -- trunk -- S2
inside LAN (hsrp)