Operating instructions

Chapter 7. Cisco Systems IGESM configuration and network integration 193
7.7 Trunk Failover feature description and configuration
This section provides an explanation of the Trunk Failover feature (available in 12.1(14)AY4
and above IOS for the IGESM) as well as several configuration examples.
For more about Trunk Failover operation and configuration, reference the IGESM Software
Configuration Guide (link provided in the online resources section later in this document).
7.7.1 Introduction to Trunk Failover
The Trunk Failover feature (also known as Link State Tracking and Layer 2 Trunk Failover) has
been available since 12.1(14)AY4 IOS on the IGESM.
The purpose of Trunk Failover is to allow a server running NIC Teaming software (combines 2
physical NICs into a single logical NIC to the OS) to know when the uplink ports out of the
IGESM go down. This prevents black holing of traffic under this condition.
Trunk Failover works by shutting down ports directly connected to the configured blade server
when the configured upstream ports go down. It does this by putting the downstream ports in
to err-disable state (down) when all upstream links in the configured group are down
In most cases you should configure Trunk Failover on all CIGESMs in the BladeCenter
connected to the bladeservers running NIC Teaming.
Some important rules for ensuring High Availability with these features
As with any network design, for high availability (HA) to be truly effective, it needs to be well
thought out. The following are some important design considerations to try to ensure
connectivity is maintained under various failure scenarios:
򐂰 For NIC teaming to work properly with Trunk Failover, you need to have external L2
connectivity between CIGESMs for proper fail-over (this is between two upstream switches
on the same L2 network/VLAN). In the designs in this document, this is achieved by
carrying the VLANs being used by the blade servers, on both IGESM’s uplink connections,
and between 6500-1 and 6500-3.
򐂰 For the combination of Trunk Failover and NIC Teaming to provide effective HA, the 6500-1
and 6500-3 in these examples must be running some sort of HSRP, and the blade servers
need to be using this HSRP address as their default gateway. If HSRP were not in use,
and 6500-1 had sole control of an upstream default gateway address, and then 6500-1
went down, Trunk Failover would see the upstream fault and drop the downstream
connections. NIC Teaming would see this and switch over to the other NIC and send the
packet toward 6500-3, but 6500-3 would not be able to get to the default gateway, and thus
drop the packet.
Figure 7-22 shows certain attributes of NIC Teaming and Trunk Failover. Note that the failure
of a NIC within the blade server, the failure of a link between the IGESM and the blade server,
and the hard failure of the IGESM (resulting in link down conditions), can all be detected by
NIC Teaming, without the aid of Trunk Failover. Trunk Failover comes into play when there is a
failure anywhere on the link between the IGESM and the upstream switch (including a hard
failure of the upstream switch that would result in a link down condition toward the IGESM).