Installation Overview

Nantian Gateway runs in Kubernetes as three workloads: a Go control plane, a Rust data plane, and an optional dashboard. Helm is the recommended install path for most users and operators. Kustomize overlays are available when you need patch-based manifests or GitOps workflows.

Installation Methods

Method	Best for	Notes
Helm	Most installations, upgrades, and value overrides.	Uses the chart repository at https://charts.nantian.dev.
Kustomize	GitOps or repository-local manifest customization.	Uses overlays from the component repositories.

Install Gateway API CRDs before installing the gateway. The quick start uses the Gateway API v1.5.1 standard CRD bundle.

What Helm Installs

A default Helm install uses the nantian-gw namespace and creates these main resources:

Resource	Default name	Default count or port	Purpose
Namespace	nantian-gw	1	Component namespace when namespace.create: true.
Control plane Deployment	nantian-gw-controlplane	2 replicas	Watches Kubernetes resources, translates them, updates status, and publishes snapshots.
Data plane Deployment	nantian-gw-dataplane	2 replicas	Receives snapshots and handles runtime traffic.
Dashboard Deployment	nantian-gw-dashboard	1 replica	Web UI for users and operators.
GatewayClass	nantian-gw	1	Uses controller gateway.networking.k8s.io/nantian-gw.

The Helm chart creates the GatewayClass; it does not create your application Gateway or HTTPRoute. Create those resources in application namespaces after installation.

Services And Ports

The current Helm templates create fixed ClusterIP services for control plane, data plane admin/metrics, and dashboard access:

Service	Port	Purpose
nantian-gw-controlplane-grpc	18080	Data plane xDS/gRPC connection.
nantian-gw-controlplane-admin	18081	Control plane admin API.
nantian-gw-controlplane-metrics	18082	Control plane Prometheus metrics.
nantian-gw-dataplane-admin	19080	Data plane admin API.
nantian-gw-dataplane-metrics	19080	Data plane metrics scrape entry.
nantian-gw-dashboard	3000	Dashboard web UI.

The data plane runtime HTTP listener is configured as 0.0.0.0:10080. Expose this listener according to your environment, for example with port-forwarding in local tests or a Kubernetes Service overlay for cluster ingress. Do not expose the control plane gRPC, admin, or metrics services outside the cluster.

Default Chart Behavior

The chart defaults are intentionally conservative:

• Images use global.imageRegistry: "ghcr.io" with repositories nantian-gw/nantian-controlplane, nantian-gw/dataplane, and nantian-gw/dashboard.
• Control plane and data plane each run two replicas by default; dashboard runs one replica.
• HPA is disabled by default with hpa.enabled: false.
• Prometheus Operator ServiceMonitor resources are disabled by default with serviceMonitor.enabled: false.
• NetworkPolicies are enabled by default with networkPolicies.enabled: true.
• The data plane has accessLogVolume.enabled: true, mounted at /var/log/nantian-gw with sizeLimit: 256Mi.
• Experimental Gateway and AI features are disabled by default in both planes.

Post-Install Checks

After installation, verify the workloads and GatewayClass:

kubectl get pods -n nantian-gw
kubectl get gatewayclass nantian-gw
kubectl get svc -n nantian-gw

If a pod is not ready, inspect recent logs:

kubectl logs -n nantian-gw deploy/nantian-gw-controlplane --tail=100
kubectl logs -n nantian-gw deploy/nantian-gw-dataplane --tail=100

Then create a demo Gateway and HTTPRoute as shown in the Quick Start.

Choosing The Next Page

• Use Helm Installation for repository setup, default values, common overrides, and uninstall commands.
• Use Experimental Features before creating AIService, TokenPolicy, WasmPlugin, or experimental BackendLBPolicy resources.
• Use Operations for first checks after install, service names, logs, metrics, and admin access.