Kubernetes-Native Network Platform
NovaEdge is a unified replacement for Envoy + MetalLB + NGINX Ingress + Cisco SD-WAN, written in Go and Rust. It combines L4/L7 load balancing, VIP management, service mesh, and SD-WAN into a Kubernetes-native platform with a Go control plane and Rust data plane.
- Protocol support: HTTP/1.1, HTTP/2 (h2/h2c), HTTP/3 (QUIC), WebSockets, gRPC, Server-Sent Events
- 8 CRD-selectable load balancing policies (RoundRobin, LeastConn, P2C, EWMA, RingHash, Maglev, SourceHash, Sticky) plus composable wrappers in the Rust dataplane (Locality-aware, Priority failover, Panic mode, Slow-start, Resource-adaptive)
- Advanced routing: hostname, path, header, method matching with boolean routing expressions
- Traffic management: canary/weighted splitting, traffic mirroring, request retry with configurable policies
- Response processing: HTTP caching, gzip/Brotli compression, buffering, custom error pages
- Composable middleware pipelines with per-route configuration
- HTTP-to-HTTPS redirect, URL rewrite, header modification (request and response)
- TCP and UDP proxying with connection tracking
- TLS passthrough (SNI-based routing without termination)
- L4 load balancing handled by NovaNet
- eBPF/XDP acceleration (enabled by default, auto-detected):
- AF_XDP zero-copy — shared-memory ring buffers for kernel-bypass packet I/O
- eBPF mesh redirect —
SK_LOOKUPsocket redirection replaces nftables/iptables TPROXY rules - SOCKMAP same-node bypass — kernel stack bypass for pod-to-pod traffic on the same node
- eBPF conntrack — efficient connection tracking via BPF maps
- Automatic fallback to legacy paths if kernel does not support eBPF
- L2 ARP mode: active-passive VIP ownership with GARP
- BGP mode: active-active multi-node ECMP via GoBGP
- OSPF mode: active-active L3 routing
- BFD (Bidirectional Forwarding Detection) for sub-second failure detection
- IPv6 dual-stack support
- IP address pools (
ProxyIPPoolCRD) with IPAM allocation
- TLS termination with SNI multi-certificate support and TLS 1.3 minimum
- Frontend mTLS (mutual TLS client certificate verification)
- OCSP stapling for certificate revocation checking
- PROXY protocol v1/v2 support
- Authentication stack: basic auth, forward auth, OAuth2/OIDC, Keycloak integration
- JWT validation with JWKS support
- Rate limiting: local (token bucket) and distributed (Redis-backed)
- CORS, IP filtering, security headers (HSTS, CSP, X-Frame-Options)
- WAF (Web Application Firewall) powered by Coraza
- Request size limits and per-route access logging
- ACME (Let's Encrypt) with HTTP-01, DNS-01, and TLS-ALPN-01 challenges
- DNS providers: Cloudflare, Route53, Google Cloud DNS
- cert-manager integration for Kubernetes-native certificate lifecycle
- HashiCorp Vault integration (PKI secrets engine, KV store)
- Self-signed certificate generation
- Certificate hot-reload without downtime
- WASM plugin system (via Wazero runtime) for custom request/response processing
- Composable middleware pipelines with boolean expressions for conditional routing
- 12 Custom Resource Definitions for declarative configuration
- Ingress API v1 support with automatic translation
- Gateway API v1 support (Gateway, HTTPRoute, GRPCRoute, TCPRoute, TLSRoute)
- Operator (
NovaEdgeClusterCRD) for lifecycle management - Multi-cluster federation with hub-spoke topology and split-brain detection
- Helm charts for controller, agent, and operator deployment
- OpenTelemetry distributed tracing
- Prometheus metrics with cardinality limiting
- Structured logging (zap) with correlation IDs
- Web UI dashboard (via novactl)
- Per-route access logging
- CLI tool (novactl) with trace query, metrics query, and agent inspection
- Transparent mTLS between services via TPROXY interception (eBPF
SK_LOOKUPauto-detected, nftables/iptables fallback) - Embedded mesh CA with SPIFFE identity (ECDSA P-256 workload certificates)
- HTTP/2 mTLS tunnel for encrypted service-to-service communication
- Mesh authorization engine with service-level access policies
- Automatic certificate rotation with configurable renewal threshold
- Multi-link WAN management with primary, backup, and load-balanced roles
- Application-aware path selection with SLA-based routing strategies (lowest-latency, highest-bandwidth, most-reliable, lowest-cost)
- WireGuard tunnels with wgctrl kernel API and STUN NAT traversal
- Real-time link quality probing -- latency, jitter, packet loss with EWMA smoothing
- Automatic failover with hysteresis to prevent link flip-flop
- DSCP marking for QoS enforcement
- 2 new CRDs:
ProxyWANLink,ProxyWANPolicy
- Dedicated VIP for controller high availability
- Health-check based failover using Kubernetes
/livezendpoint - BGP/BFD mode support for CP VIP announcement
- Active and passive health checking
- Circuit breaking with configurable thresholds
- Graceful shutdown with in-flight request tracking
- Connection pooling with per-cluster limits
NovaEdge consists of five major components:
- Operator: Manages NovaEdge lifecycle via
NovaEdgeClusterCRD - Controller (Control-Plane): Runs as a Deployment, watches CRDs and Kubernetes resources, builds routing configuration, and pushes ConfigSnapshots to node agents via gRPC
- Node Agent (Config Agent): Runs as a DaemonSet with hostNetwork, receives configuration from the controller, manages VIP binding (ARP/BGP/OSPF/BFD), iptables/nftables rules, eBPF program lifecycle, and pushes config to the Rust dataplane via gRPC. Does NOT handle user traffic.
- Rust Dataplane (Traffic Handler): Runs as a DaemonSet sidecar alongside the Go agent, binds ports 80/443, handles ALL L4/L7 traffic (HTTP/HTTPS/HTTP3/TCP/UDP/WebSocket/gRPC), executes routing, middleware, policies, load balancing, and connection pooling. Optionally accelerated by eBPF XDP.
- CRDs: 12 Custom Resource Definitions:
- Core:
ProxyGateway,ProxyRoute,ProxyBackend,ProxyPolicy,ProxyVIP - Certificate & IP:
ProxyCertificate,ProxyIPPool - SD-WAN:
ProxyWANLink,ProxyWANPolicy - Cluster management:
NovaEdgeCluster,NovaEdgeFederation,NovaEdgeRemoteCluster
- Core:
See the documentation site for detailed architecture and specifications.
- Go 1.25+
- Kubernetes cluster (1.29+)
- kubectl configured
- make
# Build all 5 binaries (controller, agent, standalone, operator, novactl)
make build-all
# Or build individually
make build-controller
make build-agent
make build-standalone
make build-operator
make build-novactl
# Build Docker images
make docker-build
# Run tests
make test
# Run tests with coverage
make test-coverage
# Run linter
make lint# Install the operator
helm install novaedge-operator ./charts/novaedge-operator \
--namespace nova-system --create-namespace
# Deploy NovaEdge via operator
kubectl apply -f - <<EOF
apiVersion: novaedge.io/v1alpha1
kind: NovaEdgeCluster
metadata:
name: novaedge
namespace: nova-system
spec:
version: "v0.1.0"
agent:
vip:
enabled: true
mode: L2
EOF
# Verify
kubectl get pods -n nova-system# 1. Install CRDs and create namespace
make install-crds
kubectl apply -f config/controller/namespace.yaml
# 2. Deploy controller
kubectl apply -f config/rbac/
kubectl apply -f config/controller/deployment.yaml
# 3. Deploy agents (DaemonSet)
kubectl apply -f config/agent/serviceaccount.yaml
kubectl apply -f config/agent/clusterrole.yaml
kubectl apply -f config/agent/clusterrolebinding.yaml
kubectl apply -f config/agent/daemonset.yaml
# 4. Verify deployment
kubectl get pods -n nova-system
kubectl logs -n nova-system -l app.kubernetes.io/name=novaedge-controller
kubectl logs -n nova-system -l app.kubernetes.io/name=novaedge-agent# Apply sample resources (NovaEdge CRDs)
kubectl apply -f config/samples/proxyvip_sample.yaml
kubectl apply -f config/samples/proxygateway_sample.yaml
kubectl apply -f config/samples/proxybackend_sample.yaml
kubectl apply -f config/samples/proxyroute_sample.yaml
kubectl apply -f config/samples/proxypolicy_ratelimit_sample.yaml
kubectl apply -f config/samples/proxypolicy_cors_sample.yaml
kubectl apply -f config/samples/proxypolicy_jwt_sample.yaml
kubectl apply -f config/samples/proxycertificate_acme.yaml
# Or use standard Kubernetes Ingress
kubectl apply -f config/samples/ingress_sample.yaml
# Or use Gateway API
kubectl apply -f config/samples/gatewayclass.yaml
kubectl apply -f config/samples/gateway_example.yaml
kubectl apply -f config/samples/httproute_example.yaml
# Check status with kubectl
kubectl get proxyvips
kubectl get proxygateways
kubectl get proxyroutes
kubectl get proxybackends
kubectl get proxypolicies
kubectl get proxycertificates
kubectl get proxyippools
# Or use the novactl CLI tool
./novactl get gateways
./novactl get routes
./novactl get backends
./novactl get vips
./novactl get policies
./novactl describe gateway my-gateway
# Advanced novactl features
./novactl agent config worker-1 # Query agent configuration
./novactl trace list --limit 20 # List recent traces
./novactl metrics query 'rate(requests[5m])' # Execute PromQL
./novactl metrics dashboard # Show metrics dashboard
# SD-WAN management
./novactl sdwan status # Show SD-WAN status summary
./novactl sdwan links # List WAN links with quality data
./novactl sdwan links -A # List WAN links in all namespaces
./novactl sdwan topology # Show overlay network topology- 1345+ test functions across unit, integration, and controller tests
- 85%+ unit coverage for critical components (router, health, VIP, load balancing, mesh, policies)
# Run all tests
make test
# Run with coverage
make test-coverage
# Run integration tests
go test -v ./test/integration/...
# Run specific component tests
go test -v ./internal/agent/router/...
go test -v ./internal/agent/health/...
go test -v ./internal/controller/...
# Run benchmarks
go test -bench=. -benchmem ./internal/agent/...- Zero linting errors with 16 strict linters enabled (golangci-lint)
- CI pipeline: gofmt, golangci-lint, go vet, govulncheck, unit tests, build all 5 binaries
- Standardized error handling with custom error types
- Interface abstractions for improved testability
Comprehensive documentation is available in the docs/ directory covering:
- Architecture: system design, component deep-dives, federation model
- User Guides: routing, load balancing, VIP management, TLS, authentication, WASM plugins, L4 proxying, WAF, and more
- Operations: observability, Web UI dashboard, troubleshooting, access logging
- Reference: CRD specifications, CLI reference, Helm values
- Development: contributing guide, development setup
See docs/development/contributing.md for contribution guidelines.
See CLAUDE.md for development guidelines when working with Claude Code.
Copyright 2024 NovaEdge Authors. Licensed under the Apache License, Version 2.0.