Skip to content

Priority-based backend groups #558

Description

@shaneutt

What

Backend priority groups that cascade failover through ordered tiers. Backends at the same priority level are load-balanced; failover cascades to the next priority group when all backends in a group are unhealthy or exhausted.

Why

Multi-provider AI deployments need deterministic failover ordering. Envoy AI Gateway's routing strategy: Primary (Priority 1, AWS Bedrock, weight 100) -> Secondary (Priority 2, GCP Vertex AI us-east5 weight 80, us-central1 weight 60). Without priority groups, all backends are treated equally regardless of cost, latency, or capacity tier.

Goals

  • Per-backend priority level in cluster configuration
  • Load balancing within a priority group (existing strategies: round-robin, P2C, etc.)
  • Cascade to next group when current group has no healthy backends
  • Integration with health checks and circuit breakers for group-level health

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Type

    Projects

    Milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions