Use the following best practices to help build your network topology:
- Enable graceful restart on your on-premises BGP device. With graceful restart, traffic between networks won't be disrupted in the event of a Cloud Router or BGP device failure as long as the BGP session is re-established within the graceful restart period.
- If graceful restart is not supported or enabled on your device, you should configure two on-premises devices with one tunnel each to provide redundancy. If you don't, Cloud VPN tunnel traffic can be disrupted in the event of Cloud Router or on-premises BGP device failure.
- For high reliability, set up redundant routers and BGP sessions even if your on-premises device supports graceful restart. In the event of non-transient failures, you'll be protected even if one path fails. For more information, see Redundant Cloud VPN tunnels.
- To connect your on-premises network to multiple Google Cloud projects by using dynamic routing, see these scenarios for VPC Network Peering or Shared VPC.