VRRP is configured on VLANs 10 and 20 which face towards the server blades and VLANs 35 and 46 which face the Core routers. VRRP uses an IP address which is shared between the two GbESM modules. At any moment, one of the GbESM modules is the VRRP master, and only the master responds to the shared address. In this sample, GbESM 1 is configured with a higher priority than GbESM 2 and it will therefore be the master if it is operational.

Hot-standby is an option of VRRP which works similarly to the trunk failover feature used with the Layer 2 configurations. When VRRP is configured with hot-standby, the standby switch will disable all the internal ports within the chassis, triggering the NIC teaming driver to use the ports which connect to the other switch.

Note: It is possible to use VRRP without the hot-standby feature. Such a design survives the failure of the primary switch module in the same way as it would with hot-standby enabled. In some circumstances the failure of the uplink ports (or upstream switch) is not protected against without the use of hot-standby. It is always advisable to test a proposed configuration to verify that it will protect against all the failure modes that you intend it to.

Merits of this configuration

This configuration provides strong High Availability features. Its behavior in response to common failure modes is as follows:

￿Switch failure, where traffic flows via the other switch; NIC teaming and VRRP ensure routing to the appropriate places. Servers default gateway is still valid and static routes pointing to the externally facing VRRP addresses are still valid.

￿Uplink failure, where traffic flows either via the other switch and through a cross-over link connecting the two GbESM switches or, with hot standby and tracking, disables the internal ports and trigger NIC teaming thus sending all traffic through the second switch.

Summary of disconnect procedure to be performed for each example

When performing initial configurations or making changes to existing configurations that might have an impact on a routed network, it is recommended that you leave connections un-cabled, or shut down, prior to making the configuration changes. This will reduce the likelihood of any routing loops or other transient behaviors which might disrupt other parts of the network.

￿Shut down the ports on GbESM1 and GbESM2

/oper/port EXT1/dis /oper/port EXT2/dis /oper/port EXT5/dis /oper/port EXT6/dis

￿Shut down the ports on Core1 and Core2

conf t

int range g0/1-2, g0/11-12 shut

110Nortel Networks L2/3 Ethernet Switch Module for IBM Eserver BladeCenter

Page 124
Image 124
IBM L2/3 manual Merits of this configuration