Skip to main content
LKS exposes Workloads to the Internet through Kubernetes Service objects of type LoadBalancer. Each LoadBalancer Service is backed by Latitude Elastic IPs via BGP and MetalLB: it’s assigned an Elastic IP that is announced from your Cluster over BGP, so traffic to that address reaches your Nodes.

​
How it works

  • L4 load balancing. LoadBalancer Services operate at layer 4 (TCP and UDP). There’s no managed HTTP layer.
  • Elastic IP over BGP. When you create a LoadBalancer Service, an Elastic IP is allocated in the Cluster’s region and MetalLB announces it over BGP, then updates the Service with the address.
  • Bring your own ingress. For HTTP routing, TLS termination, or path and host rules, run your own Ingress or Gateway controller (such as Traefik or a Gateway API implementation) behind a LoadBalancer Service. LKS doesn’t install one.

​
High availability

For a Service to stay reachable when a Node is lost, run at least two replicas of its backend, spread across Nodes. Set externalTrafficPolicy: Local on the Service. Under this policy the IP is announced only from Nodes that run a backing Pod, which preserves the client source IP and routes traffic straight to a local Pod. With a single replica, the address is announced from just one Node, so the Service goes down if that Node or Pod fails. Two or more replicas across Nodes keep it announced from several Nodes.

​
Configure a Service

For the required Service fields and the supported annotations, see LoadBalancer Services.