Automatic fast-leave IGMP

IGMP fast-leave is configured for ports on a per-VLAN basis. By default, the switches send IGMP Group-Specific Query message out of the interface, upon which the Leave Group message is received to ensure that no other receivers are connected to the interface. If receivers are directly attached to the switch, there is no point in sending the membership query as the receiver wanting to leave is the only connected host.

Fast-leave processing eliminates the IGMP Group-Specific Query message. Thus, it allows the switch to immediately remove an interface from the bridge table upon receiving the Leave Group message. This processing speeds up the overall leave process and also eliminates the CPU overhead of having to generate an IGMP Group-Specific Query message.

Depending on the switch model, fast-leave is enabled or disabled in the default configuration.

With fast-leave enabled and an IGMP Group Leave being received on a noncascaded port, the following events take place:

  • The switch stops forwarding multicast traffic for that group to that port.

  • Does not apply to cascaded ports.

When disabled or when the port is cascaded, the regular IGMP leave time is used (up to 10 seconds when the switch is not the IGMP Querier).

On switches that do not support data-driven IGMP, unregistered multicast groups are flooded to the VLAN rather than pruned. In this scenario, fast-leave IGMP can actually increase the problem of multicast flooding by removing the IGMP group filter before the Querier has recognized the IGMP leave. The Querier will continue to transmit the multicast group during this short time, and because the group is no longer registered, the switch will then flood the multicast group to all ports.

On switches that do support data-driven IGMP ("Smart" IGMP), when unregistered multicasts are received the switch automatically filters (drops) them. Thus, the sooner the IGMP leave is processed, the sooner this multicast traffic stops flowing.

Because of the multicast flooding problem mentioned above, the IGMP fast-leave feature is disabled by default on all switches that do not support data-driven IGMP (see Operation with or without IP addressing). The feature can be enabled on these switches via an SNMP set of this object:

hpSwitchIgmpPortForceLeaveState.<vid>.<port number>
    

However, Hewlett Packard Enterprise does not recommend this because it will increase the amount of multicast flooding during the period between the client's IGMP leave and the Querier's processing of that leave. For more information on this topic, see Forced fast-leave IGMP.

If a switch port has the following characteristics, the fast-leave operation will apply:
  • Connected to only one end node.

  • The end node currently belongs to a multicast group, that is, is an IGMP client.

  • The end node subsequently leaves the multicast group.

Then the switch does not need to wait for the Querier status update interval, but instead immediately removes the IGMP client from its IGMP table and ceases transmitting IGMP traffic to the client. (If the switch detects multiple end nodes on the port, automatic fast-leave does not activate—regardless of whether one or more of these end nodes are IGMP clients.)

In the following figure, automatic fast-leave operates on the switch ports for IGMP clients "3A" and "5A," but not on the switch port for IGMP clients "7A" and "7B," server "7C," and printer "7D."

Figure 1: Example of automatic fast-leave IGMP criteria

When client "3A" running IGMP is ready to leave the multicast group, it transmits a Leave Group message. Because the switch knows that there is only one end node on port A3, it removes the client from its IGMP table and halts multicast traffic (for that group) to port A3. If the switch is not the Querier, it does not wait for the actual Querier to verify that there are no other group members on port A3. If the switch itself is the Querier, it does not query port A3 for the presence of other group members.

Fast-leave operation does not distinguish between end nodes on the same port that belong to different VLANs. Thus, for example, even if all the devices on port A6 shown in figure 1 belong to different VLANs, fast-leave does not operate on port A6.