The Fast-Leave Trap 678
The Fast-Leave Trap
Fast-Leave is a great tool in an organization that uses quite a bit of multicasting. There can be
a problem though, when using it in a network where spanning tree changes frequently.
When a switch configured for Fast-Leave receives a Leave message, the switch will remove the
port at which the message arrived from the forwarding table for the particular stream. What
happens if this occurs on a core switch, on the port going out to a closet switch or stack? The
core switch will remove all entries associated with that port. If several clients were listening to
the stream and one leaves, the core switch will remove them all.
Whenever possible, only enable Fast-Leave processing on switches that have clients terminating
at individual ports. Turn this feature on at the closets, but think twice before doing so at the core
and distribution layer switches.
Multicast forwarding is relatively new. Until the growth of applications that required multicast
delivery, it was used by service protocols such as “all OSPF routers.” Now, many multimedia
applications—such as video—and wide distribution applications—such as market data feeds—
all require multicast delivery.
We have therefore given a lot of time to understanding the many facets of IP multicast.
We started with an overview of multicast and compared it to unicast and broadcast communication,
and then discussed how IP addresses were designated as multicast addresses.
These layer 3 addresses must be converted to layer 2 MAC addresses using a standard mapping
Of course, theoretical knowledge needs to be backed up with an understanding of how to
configure multicast on both Cisco routers and switches, because the routers carry the multicast
traffic over the internetwork and the switches deliver it to the multicast hosts. The syntax of the
commands is straightforward, but you need to ensure that the network is properly planned
before starting the implementation. Care needs to be taken when considering the IGMP version
to be used, for example.
When considering the multicast distribution, you need to select between PIM DM, PIM SM,
and CBT. All three are independent protocols that use tree distribution to manage multicast
data delivery in a network, but all three affect the network operation in different ways and
require different configurations.
Know the difference between IGMP and IGMPv2 and IGMPv3. IGMPv1 and IGMPv2 are
very similar. The major difference is that IGMPv2 has a message that the client sends when it
doesn’t want to receive the multicast stream anymore. The result of this small difference is that
they don’t work well together, and it makes sense for you to have only one version running.
IGMPv3 is better still and supports extras such as SSM, but because it is very new, you have to
make sure that all the hosts in your network support the Leave message
Exam Essentials 635
The Catalyst switch can listen for client Leave messages with both CGMP and IGMP Snooping.
A switch configured for CGMP can listen for IGMPv2 and v3 Leave messages by being configured
with the command set cgmp leave enable. A switch configured for IGMP Snooping can be
configured for IGMP Fast-Leave processing with the command set igmp fastleave enable.
Know the difference between CGMP and IGMP Snooping. Although both CGMP and IGMP
Snooping allow a switch to get involved in a multicast stream, they are very different protocols.
CGMP is a Cisco proprietary protocol communication based on communication between a router
and any attached switches. Routers receiving IGMP packets from other routers forward specific
information to appropriate switches containing information on multicast memberships.
IGMP Snooping enables the switch to learn information from watching IGMP packets go
through the switch. IGMP is an Internet standard. Snooping is being considered by the IETF as
a standard and is currently in draft. Remember that snooping can’t be enabled if CGMP is
enabled, so you first need to make sure that CGMP is turned off. Next, enable IGMP Snooping
with the ip igmp snooping command.
Know the difference between the multicast routing protocols. There are several options for
routing multicast traffic. DVMRP is a distance-vector-based routing protocol, and MOSPF uses
OSPF, but neither is the recommended method of doing multicast routing with Cisco equipment.
Cisco recommends that PIM be used to route multicast streams because it learns from the preexisting
routing protocol. This means that EIGRP can be used to route multicast information.
PIM has two broad modes: sparse and dense. In dense-mode PIM, each router is automatically
included in the multicast table and has to prune itself off if no clients need the stream. Sparse mode
assumes no routers wish to participate. Routers are added as connected clients request access to
the multicast streams, and a special router is used as the base for the entire tree. This router—a
rendezvous point—needs to be referenced in each multicast router’s configuration. Use the command
rp pim ip-address ip_address to define the IP address of the rendezvous point.
Know how to troubleshoot your multicast setup. There are many show commands that can
be done on the router and switch to show communication, but you still need to test the transport.
You can use the ping command to reach out and touch a particular multicast IP address.
If you want to do a traceroute, use the command mtrace
410 times read