How "loopback keepalive" (Ether 0x9000) work?

Konstantin Dunaev · ‎06-02-2007

Hi,

I met quite often the opinion that Cisco's implementation on Ethernet loopback keepalives (Ether type 0x9000) make life difficult when your STP doesn't converge fast enouch. We had such a bad experience when all our switches put their uplinks in Err-disable status after one port was wrong connected and STP didn't converege before "looopback" was detected.

How exactly does the loopback keepalive work? I know that it sends the Ethernet frame out of a port with DEST and SRC MAC address of that interface. According to normal logic the upstream switch will forward that ethernet frame back to the original interface and the first switch will get their own frame back.

But in which case the loopback condition will be detected?

Konstantin Dunaev · ‎06-02-2007

What will happen to this loopback keepalive frame if it's recieved by a switch, which has emty MAC table (eg. all MAC address was flashed after recieving the STP change notification)?

will it be flooded over all ports?

Francois Tallet · ‎06-04-2007

No, a switch will not send back a frame the port on which it has received it, irrespective of its cam table. Well, at least not in the same vlan. I think the problem you had is that your traffic was looped back on the same port but on a different vlan. I'm trying to determine whether it would trigger the errdisable mechanism. Looks like it does but want to double check.

F.

Konstantin Dunaev · ‎06-04-2007

hm, exactly, I remember something like this, that switch should not send the frame back to the port, where it was got from. I was confused a little bit because in the Net they said that this keepalive message is used by device to check the availability of connection - if it sends the loopback keepalive fram and gets answer then the connnection is OK.

But with different VLANs it would be interestin to know.

and, what is your opinion, should we disable the "keepalive" on our uplinks?