feat: Service mesh integration (Istio/Linkerd)

[gouravjshah]

Summary

Add support for service mesh integration to enable mTLS, traffic policies, and enhanced observability.

Parent Epic

Part of #1 - Production Kubernetes & Container Support

Motivation

Service meshes provide:

• mTLS: Automatic encryption between services
• Traffic management: Retries, timeouts, circuit breaking
• Observability: Distributed tracing, golden metrics
• Security: Authorization policies

Istio Integration

Sidecar Injection

# Helm values for Istio
podAnnotations:
  sidecar.istio.io/inject: "true"

Virtual Service

apiVersion: networking.istio.io/v1beta1
kind: VirtualService
metadata:
  name: mcp-gateway
spec:
  hosts:
    - mcp-gateway
  http:
    - route:
        - destination:
            host: mcp-gateway
      timeout: 30s
      retries:
        attempts: 3
        perTryTimeout: 10s

Authorization Policy

apiVersion: security.istio.io/v1beta1
kind: AuthorizationPolicy
metadata:
  name: mcp-gateway
spec:
  selector:
    matchLabels:
      app: mcp-gateway
  rules:
    - from:
        - source:
            principals: ["cluster.local/ns/ai-platform/sa/claude-agent"]

Linkerd Integration

Annotation for injection

podAnnotations:
  linkerd.io/inject: enabled

Service Profile

apiVersion: linkerd.io/v1alpha2
kind: ServiceProfile
metadata:
  name: mcp-gateway.default.svc.cluster.local
spec:
  routes:
    - name: mcp-endpoint
      condition:
        pathRegex: /servers/[^/]+/mcp
      responseClasses:
        - condition:
            status:
              min: 500
          isRetryable: true

Features Required

• Documentation for Istio integration
• Documentation for Linkerd integration
• Helm chart options for mesh annotations
• Virtual Service / Service Profile templates
• mTLS configuration options
• Distributed tracing headers support (x-request-id, etc.)

Acceptance Criteria

• Gateway works with Istio sidecar
• Gateway works with Linkerd proxy
• mTLS enabled between services
• Traces appear in Jaeger/Zipkin
• Authorization policies work correctly
• Documentation covers both meshes

References

• Istio documentation
• Linkerd documentation
• Service mesh comparison