[j-nsp] Juniper EX, HP server NIC teaming

[Dale Shaw]

Hi all,

I'm operating in a bit of an information vacuum here, but I'm trying
to help out some colleagues with a server NIC teaming / EX switch
problem.

The problem as described to me is that some HP servers with teamed
NICs are having connectivity problems when migrating from Cisco to
Juniper LAN switches. The teaming mode on the server side is "Network
Fault Tolerance (NFT)". Theoretically this means only one adapter in
the team is active at any one time, however there is a heartbeat
protocol used to detect faults in the switching fabric. The heartbeat
protocol is based on a Layer 2 multicast -- it is not IP. The source
MAC is the 'real' adapter's hardware address, and the destination MAC
is 03-00-C7-00-00-EE.

The server appears to flip flop between interfaces in the team, which
indicates to me that something is going wrong with the heartbeat.
We're about to do some lab work to see what's happening on the wire
but in the meantime I thought I'd shoot out this e-mail.

My colleagues connected the server to a Cisco switch which had been
"write erase, reload"'d, and it just worked.

Is there anything "special" about the way a Juniper EX switch handles
non-IP, multicast frames? The EX switch config is really, really
simple. They've tried turning just about everything off (including
STP) to no avail.

Any clues? Did some Googling but came up empty.

cheers,
Dale
BTW: Having said all that about heartbeats, I could be headed down a
rabbit warren -- if the high level problem description reminds any of
you of another problem, please set me straight :-)