Greetings,
We've got the following setup:
[ other AS ]
|
[ mlx-4 ] -- [ icx6430 ] -- (cloud) -- [ customer ]
|
[ server ]
The mlx has a bgp session with the other AS and with the customer. The
server and other AS can reach each other, the customer and the server
can reach each other and the other AS and customer can reach each other.
This setup is duplicated many times, with different servers and
different customers.
From time to time a server gets unreachable from the customer and vice
verse. The other AS can still reach the server and the customer.
I made a port-mirror of the ports on the mlx-4 and the icx6430 that face
each other and captured the data going through. Although there's nothing
between the two devices, except for a direct patch, I see packets on the
icx, that I don't see on the mlx. The packetloss is not random, so it's
not a problem in the cable. When I ping the customer from the other AS
and the server, I see the reply packets for both on the icx, but only
for the other AS on the mlx.
So, the problem isn't at the customer, the routing on their side still
works. Somehow the mlx is dropping packets, but I have no idea why. It's
not always the same customer where the problem occurs and we couldn't
find anything that triggers this event.
Now the weirdest part. To "fix" the problem, it helps to change the
mac-address of the server. In some cases it also helped to do 'show
mac-address' on the mlx.
Does anyone have an idea where to look?
Thanks in advance,
Rogier
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
We've got the following setup:
[ other AS ]
|
[ mlx-4 ] -- [ icx6430 ] -- (cloud) -- [ customer ]
|
[ server ]
The mlx has a bgp session with the other AS and with the customer. The
server and other AS can reach each other, the customer and the server
can reach each other and the other AS and customer can reach each other.
This setup is duplicated many times, with different servers and
different customers.
From time to time a server gets unreachable from the customer and vice
verse. The other AS can still reach the server and the customer.
I made a port-mirror of the ports on the mlx-4 and the icx6430 that face
each other and captured the data going through. Although there's nothing
between the two devices, except for a direct patch, I see packets on the
icx, that I don't see on the mlx. The packetloss is not random, so it's
not a problem in the cable. When I ping the customer from the other AS
and the server, I see the reply packets for both on the icx, but only
for the other AS on the mlx.
So, the problem isn't at the customer, the routing on their side still
works. Somehow the mlx is dropping packets, but I have no idea why. It's
not always the same customer where the problem occurs and we couldn't
find anything that triggers this event.
Now the weirdest part. To "fix" the problem, it helps to change the
mac-address of the server. In some cases it also helped to do 'show
mac-address' on the mlx.
Does anyone have an idea where to look?
Thanks in advance,
Rogier
_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp