P2-11 — Advanced backend configuration: multiple backends and load-balancing modes

[bihius]

Summary

Extend the HAProxy config generation to support richer, real-world deployment topologies that go beyond the current single-backend MVP model.

Motivation

The current implementation generates a fixed single-backend HAProxy configuration. Production deployments commonly require:

• Multiple named backend pools (e.g. split by service, tenant, or region)
• Different load-balancing algorithms per backend
• Health-check tuning per backend

Proposed scope

Multiple backends

• Allow the admin to define more than one backend pool, each with its own set of upstream servers and settings.
• The config generator must produce a separate backend block for each pool and wire the correct ACL / use_backend rules in the frontend.

Load-balancing modes

Support at minimum the following HAProxy algorithms, selectable per backend:

Algorithm	HAProxy keyword
Round-robin (default)	roundrobin
Least connections	leastconn
Source IP hash	source
URI hash	uri
Random	random

Health checks

• Configurable check interval, rise, and fall thresholds per backend.

Out of scope (for this issue)

• Sticky sessions / cookie-based persistence (separate issue if needed)
• mTLS between HAProxy and upstreams

Acceptance criteria

• Data model extended to store multiple backends per proxy rule
• Config generator produces syntactically valid multi-backend HAProxy config (haproxy -c passes)
• Admin UI allows creating/editing backends and selecting the load-balancing algorithm
• Unit tests for new config generation paths
• Integration tests for new API endpoints