On Fri, Jun 30, 2000 at 08:44:00AM -0700, Brad Knotwell wrote:
> Hello all--
>
> I've been having a fascinating problem involving spurious arp requests
> (fairly vanilla 2.2.16 kernel). I have multiple ethernet cards serving
> separate subnets (FWIW: 192.168.100 and 192.168.200; eth0 and eth1
> respectively). The 100 net is served through a load-balancer that reaps
> connections if they are idle for too long.
>
> Unfortunately, after these connections are reaped, it *appears* (I
> say appears because I only suspect the reaping causes the arp) Linux
> periodically generates an arp on the 100 net using the MAC address of
> eth1 (200 net;the "tell" refers to the 100 net address). As you might
> suspect, the load balancer receives this arp, updates its arp table for
> that address, and connectivity over the 100 net (the 200 net still works) is
> lost between these two machines until another arp request (this time
> referencing the MAC address associated with the card on the 100 net)
> occurs.
Yes, Linux ARP does regular unicast ARP probing when it suspects a dead
neighbour. See man 7 arp for the details.
-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
