Load Balancers

Hubfly space Load Balancers distribute public traffic across multiple backend container targets behind a single custom domain or managed endpoint. Use them to scale application request throughput, distribute load across replicas, or manage blue/green traffic shifts.

Load balancers are created from the project overview's Create menu, which opens the Topology Studio editor — a single dialog for naming the balancer, wiring ingress domains on the left, and adding weighted backend targets on the right.

Balancing Algorithms

Pick a strategy from the Studio editor's Strategy dropdown when configuring the balancer node:

The Create Load Balancer Topology Studio dialog with the Strategy dropdown open, showing Round Robin, Least Connections, and IP Hash options — The Studio editor's Strategy dropdown — Round Robin, Least Connections, and IP Hash, set on the load balancer node itself.

Algorithm	Routing Strategy	Best Use Case
`round_robin`	Distributes inbound requests sequentially across active target containers.	Stateless web applications and REST APIs with uniform hardware capacity.
`least_conn`	Routes new connections to the target container with the lowest active connections.	Long-lived HTTP connections, WebSockets, or variable processing workloads.
`ip_hash`	Binds client requests to the same target container based on client IP.	Workloads requiring sticky client sessions or local cache affinity.

Target Weights & Zero-Downtime Rollouts

Assign integer weights (e.g. 1 to 10) to backend targets to perform weighted traffic distribution.

Canary Release Pattern

Deploy api-v2 next to existing api-v1 targets.
Add api-v2 to the load balancer group with a low weight (e.g. weight: 1 vs weight: 9).
Monitor error rates and latency logs on api-v2.
Gradually increase weight to 100% and detach api-v1.

Health Checks & Target Isolation

Load balancers evaluate target readiness using lightweight health check probes (e.g. /health or /ready). If a target container fails health checks, it is automatically removed from the active routing pool until it recovers.