Customer Edge high availability

F5 Distributed Cloud Customer Edge (CE) high availability can exist at two different layers. Choose the layer before you choose a routing protocol: several independent CE sites provide routing-level availability, while the nodes of one CE site provide site-level clustering.

Design	Failure unit	Traffic distribution	Use it when
Independent CE sites selected by a Virtual Site	One CE site	The upstream router installs equal-cost Border Gateway Protocol (BGP) paths to the sites	The network supports BGP and Equal-Cost Multi-Path (ECMP), and the application does not require a shared local virtual IP address (VIP) implemented by the CE nodes
One multi-node CE site	One node within the site	The cluster owns the site’s local forwarding behavior; Virtual Router Redundancy Protocol (VRRP) can provide a local next hop where supported	The attached network requires a locally redundant gateway or local load-balancing behavior that independent sites do not share

These designs solve different problems. A Virtual Site groups sites for policy and service selection; it does not turn them into one cluster. A cluster groups nodes into one site; it does not make an upstream router distribute traffic across separate sites.

Prefer routing-level HA when the network can withdraw failed paths

With routing-level HA, each CE site advertises the same service prefix. The upstream router installs equal-cost paths and distributes new flows across the sites. When one site’s path is withdrawn, the router stops selecting that path after the routing protocol converges.

This pattern scales by adding sites, but the effective fan-out is bounded by the upstream router. If the router supports fewer equal-cost next hops than the number of healthy CE sites, some advertised paths will not carry traffic. Validate the router’s ECMP limit and behavior before selecting a site count; there is no portable universal CE count.

Use clustering when the local network requires it

VRRP elects an active router from multiple routers that share a virtual address. A single-node CE site has no second node to elect and therefore cannot provide VRRP redundancy. Where the F5 Distributed Cloud site type and environment support VRRP, create the required multi-node site rather than describing a single node as a cluster.

High Availability is a site-creation decision

For Secure Mesh Site v2, enabling High Availability changes the site from one node to the required multi-node form. The setting cannot be changed after the site is created and deployed. Confirm the current node-count and platform constraints in Create Secure Mesh Site v2 before building the site.

Do not combine the names of the two designs

Use single-node site for one independently managed CE and multi-node cluster for one site whose nodes participate in site-level HA. Calling an independent set of single-node sites a cluster hides the important failure and routing boundaries.

Routing and failoverAdvertise a service prefix, remove failed paths, and plan for ECMP limits.

ECMP affinityUnderstand flow hashing and why it is not application persistence.

Standards references

• RFC 5798 defines VRRP and its election among multiple routers.
• Secure Mesh Site v2 FAQs describe current site constraints that affect node and interface design.