15.4.5 Continue multicast communication in case of master switch failure

IPv4 multicast supports the ability to continue multicast communication of PIM-SM and PIM-SSM when the master switch fails in stacked operation and the backup switch becomes a new master switch. This function is enabled when the configuration command ip pim nonstop-forwarding is set.

This function applies to multicast forwarding entries that have switched to the shortest path before the master switch fails. Therefore, even if this function is used, packet loss occurs during normal PIM-SM processing (such as switching the shortest path through the rendezvous point).

After a failure occurs in the master switch, multicast communication continues with the hardware entry set before the failure occurs within the re-learning time (330 seconds). Multicast forwarding entries that were not learned within the relearning time are deleted when the relearning time ends. An operation message is output at the start and end of the relearning time.

When this functionality is used, this functionality is enabled on all interfaces of VRF that use IPv4 multicast if IPv4 multicast is used on VRF interfaces.

You can use the following operation commands to check the relearning status of IPv4 multicast forwarding entries after an error:

show ip mcache (show ip pim mcache)
show ip mroute

<Structure of this section>

(1) Prerequisites for Continuing Multicast Communication

(a) Using Graceful Restart for Unicast Routing

To continue multicast communication when the master switch fails, enable graceful restart of the unicast routing protocol. If graceful restart is disabled, multicast communication packet loss may occur frequently because the upstream unicast path is not stable after the failure.

(b) Installing Devices Supporting Optional Generation ID

Install a device that supports Generation ID option (in accordance with RFC4601 and RFC5059) in the neighboring router of the Switch. If the neighboring router does not support Generation ID option, relearning may not be completed within the relearning period due to delays in sending and receiving PIM messaging after a failure. For details about Generation ID options, see (3) Neighbor Discovery in 15.4.2 IPv4 PIM-SM.

(2) Timer Specifications for Continuing Multicast Communication

When this function is enabled, the following timer specifications are changed.

Join/Prune-Holdtime

When this function is used with a Switch that sends PIM-Join/Prune messages upstream, Join/Prune-Holdtime is sent with the larger of the following values:

Join/Prune-Period(30~3600 seconds) NG 3.5
Join/Prune-Period(30~3600 second + re-learning time (fixed to 330 seconds)

PIM-Candidate-RP-Advertisement messaging RP-Holdtime

When this function is used as a rendezvous point candidate, PIM-Candidate-RP-Advertisement messaging RP-Holdtime is set to 290 seconds and advertised.

PIM-Bootstrap messaging RP-Holdtime

When this function is used with the bootstrap router, if PIM-Candidate-RP-Advertisement message reported RP-Holdtime is less than 165 seconds, PIM-Bootstrap message RP-Holdtime is set to 165 seconds and sent (when this function is not used, RP-Holdtime reported by PIM-Candidate-RP-Advertisement message is set).

(3) Switching the upstream interface within the re-learning time

If the unicast route of the upstream interface of a multicast forwarding entry is changed during the re-learning period, the upstream interface of the corresponding entry is also changed according to the unicast route in the same way as normal. At this time, packet loss occurs temporarily.

To suppress this unicast route change, configure a static route for the unicast route to the source, or set keep-incoming parameter using ip pim nonstop-forwarding of the configuration command. If a unicast route change causes the upstream interface to change to the downstream interface of the corresponding entry, forwarding to this downstream interface is stopped.

(4) Relay operation when the configuration command is changed

(a) Changing the Address-Range Used by PIM-SSM

If the address-range used by PIM-SSM is changed by the configuration command within the relearning time, the relearning time ends. In this case, the communication of unreleased multicast forwarding entries is stopped within the re-learning period.

(b) Configuring and Removing IPv4 Multicast for VRF

If IPv4 multicast of VRF is set or deleted by using a configuration command within the re-learning period, the re-learning period ends. In this situation, all the remaining VRF stop communication for multicast forwarding entries that have not been learned within the re-learning period.

(5) Notes

When the re-learning time ends, the non-communication of multicast forwarding entries is not monitored, and the unreleased multicast forwarding entries are deleted. Therefore, even if the entry hold time during non-communication is set longer than the relearning time by the configuration command ip pim keep-alive-time, the corresponding multicast forwarding entry is deleted when the relearning time ends.
Multicast forwarding entries in multicast extranets are not subject to this function. Therefore, if a failure occurs in the master switch, multicast communication of the corresponding entry is stopped. After an error occurs, communication resumes when the corresponding entry is relearned.
If double forwarding of multicast packets occurs immediately before the master switch fails, it may take time to resolve the duplicate forwarding. If multicast routing information is relearned after an error occurs, duplicate forwarding is cleared by PIM-Assert messaging.
If this function and the configuration command ip pim multiple-negative-cache are set at the same time, ip pim multiple-negative-cache command setting is ignored. If this happens, a warning message is displayed on the operation command show ip mcache (show ip pim mcache).
When the master switch is switched, the new master switch remains in the switching status until relearning is completed (the switch status is indicated by (Switchover) in the operation command show switch).