Traffic Management

Configurations in Traffic Management.

Configurations

• Authority Override — Demonstrates overriding the HTTP Host header sent to a specific upstream cluster, including requests received over HTTP/2
• Basic Reverse Proxy — A minimal HTTP reverse proxy with one listener and one upstream cluster.
• Bound Upstream Condition — Gates a filter on the logical upstream the router bound for the request
• Bound Upstream Dispatch — Selects an upstream endpoint straight from the logical binding, with no second router
• Canary Routing — Sends 10% of traffic to a canary backend while the stable backend handles the remaining 90%
• Circuit Breaker — Prevents cascading failures by tracking consecutive upstream errors per cluster
• Cluster Application Metadata — Tags an upstream cluster with opaque application metadata that consuming filters interpret
• Endpoint Selector — Selects an upstream endpoint from a trusted mutation source (e.g. external processing)
• Grpc Detection — Detects gRPC requests from the content-type header and promotes the variant to filter metadata and results
• Grpc Timeout — Honours the grpc-timeout request header as a real deadline
• Health Checks — Per-cluster health checks probe endpoints on a timer and remove unhealthy backends from the load balancer rotation
• Hostname Upstream — Demonstrates using DNS hostnames instead of IP addresses for upstream endpoints
• Hosts — One listener serves multiple domains
• Iterative Request Router Failover — Demonstrates provider failover using the iterative_request_router filter
• Iterative Request Router Origin Failover — Demonstrates origin-aware failover: only an upstream 429 (rate limit) triggers the fallback
• Least Connections — Routes each request to the backend with the fewest in-flight requests
• Maglev — Pins a user’s requests to one backend by hashing a request header through a Maglev lookup table
• P2C — Samples two random endpoints and picks the one with fewer in-flight requests
• Path Based Routing — Routes by URL path prefix
• Priority Lb — Defines primary and failover endpoint tiers
• Random — Selects an upstream endpoint at random, weighted by endpoint weight
• Rate Limiting — Token bucket rate limiter with per-IP or global modes
• Redirect — Returns a 3xx redirect without contacting any upstream
• Retry Policy — Automatically retries failed upstream requests with exponential backoff and a token-bucket budget to prevent retry storms
• Ring Hash — Routes requests to a stable backend by hashing a configurable request header through a sorted virtual-node ring
• Round Robin — Default strategy Configuration example for praxis.
• Session Affinity — Hashes a request header to pin a user’s requests to one backend
• Static Response — Returns a fixed response without contacting any upstream
• Sticky Sessions — Pins clients to a specific backend across requests
• Subset Lb — Filters endpoints by metadata labels and applies an inner strategy within the matching subset
• Timeout — Returns 504 if the upstream takes longer than timeout_ms to respond
• Weighted Load Balancing — Traffic split proportional to per-endpoint weights
• Zone Aware — Prefers same-zone endpoints to reduce cross-zone network costs and latency