The design in this guide uses one DCI switch at each site. If the DCI switch fails at Site 1 or Site 2, all VMs in the site that is down are automatically brought up in the other site. If the DCI switch fails at Site 3, the witness host loses connectivity, but the VMs in the stretched cluster continue to run in their existing sites. In a production environment, consider using a pair of DCI switches at each site for increased redundancy.

This chapter details the configurations of the three DCI switches in this deployment example. Alternate DCI switch configurations may be used with the following considerations:

• BGP must be properly configured between the DCI and SmartFabric switches.
• Traffic must be routed properly between all switches.
• All DCI switch ports that pass vSAN traffic must have the MTU set to at least 9216 bytes. This includes the ports connected to the witness host.
• There may be additional switches located between the DCI sites, but latency requirements specified in the vSAN Planning and Deployment Guide must be met end to end.

The DCI switch topology with configuration details is shown in the following figure.

The figure includes port numbers, port channel (Po) numbers, ASNs, and router IDs used in the switch configuration steps for reference.

Point-to-Point IP networks are configured between switches for BGP. The IP addresses in the following table are used in the switch configurations. The IP addresses are examples only. Use addresses suitable for your environment.