[evla-sw-discuss] Igloo and dropped mcast packets

[James Robnett]

   I think I figured out why igloo was dropping those mcast
packets.

   While ironing out the plans for the new 1Gbit VLA-AOC link
I discovered that igloo isn't directly connected to the
146.88.201.xxx master switch.  It goes through a small switch
in the correlator room we needed for a few extra 146.88.201
drops in that room.

    The mcast traffic is being forwarded from a 1Gbit interface
across to the 146.88.201 switch and then to that small 100Mbit
switch.   My guess is that little switch isn't updating  mcast
requests properly.  Back when we had one massive mcast group
it never timed out but with the new fragmented list it wasn't
updating in time.

    If we move igloo to the 10.64.100.xxx network or directly
off the master 146.88.201.xxx switch the problem should go away.
I'd prefer the former if igloo is going to become a primary
part of EVLA.

    I'd suggest changing it's addressing when it's upgraded
to RHEL5.3.

James