Example 2a: HP ProCurve 2626 (H.07.50) snooping IGMP v3 with two receivers

It appears that the switch totally bolloxes this up in a manner similar to the bug with release H.07.31. When host 10 leaves, it is not immediately pruned by the switch but does eventually time out in about 2 minutes. When host 11 leaves the membership report to the router is suppressed, the switch times out after about 2 minutes, and then the multicast traffc floods all ports for another minute, returning on host 10's port where it had gone away when the switch first pruned host 10.

Since I've just become aware of a whole new release train for this switch, I'm going to retry this experiment with that train before wasting a lot of time cutting and pasting tcpdump output:-) See the raw traces.

The switch

The switch under test is an HP 2626 ProCurve running version H.07.50 with IGMP snooping enabled:

; J4900A Configuration Editor; Created on release #H.07.50

hostname "HP ProCurve Switch 2626" 
cdp run 
ip default-gateway 128.59.7.1 
snmp-server community "public" Unrestricted 
vlan 1 
   name "DEFAULT_VLAN" 
   untagged 1-26 
   ip address 128.59.7.254 255.255.255.0 
   ip igmp 
   exit 
password manager

The traces

Raw tcpdump traces and console logs are here:

host 10
host 11
host 12
host 13
switch
router