Back to blog Cilium Service Mesh: Sidecar-Free with eBPF
· 6 min read · - views

Cilium Service Mesh: Sidecar-Free with eBPF

K8sNetworking
#cilium#service-mesh#ebpf#kubernetes#mtls#networking

Cilium Service Mesh: Sidecar-Free with eBPF

Traditional service meshes inject sidecar proxies into every pod. Cilium does it differently - eBPF programs in the kernel handle mTLS, load balancing, and observability with zero sidecars.

This guide covers deploying Cilium service mesh and configuring traffic management, security policies, and observability.

Cilium logo

TL;DR

  • Cilium mesh = eBPF-powered service mesh, no sidecars
  • Per-node Envoy for L7 processing (not per-pod)
  • Native mTLS with SPIFFE identities
  • Hubble for observability
  • 50% less resource overhead vs sidecar mesh

Why Sidecar-Free?

SIDECAR MESH (Istio/Linkerd)        CILIUM MESH
========================            ===========
Pod 1: App + Envoy (150MB)          Pod 1: App only
Pod 2: App + Envoy (150MB)          Pod 2: App only  
Pod 3: App + Envoy (150MB)          Pod 3: App only
                                    Node: Cilium Agent + Envoy
                                    
Memory: 450MB sidecars              Memory: ~200MB per node
Latency: 2 proxy hops               Latency: kernel-level

Install Cilium with Service Mesh

# Install Cilium CLI
CILIUM_CLI_VERSION=$(curl -s https://raw.githubusercontent.com/cilium/cilium-cli/main/stable.txt)
curl -L --remote-name-all https://github.com/cilium/cilium-cli/releases/download/${CILIUM_CLI_VERSION}/cilium-linux-amd64.tar.gz
sudo tar xzvfC cilium-linux-amd64.tar.gz /usr/local/bin

# Install with service mesh features
cilium install \
  --version 1.15.0 \
  --set kubeProxyReplacement=true \
  --set ingressController.enabled=true \
  --set ingressController.loadbalancerMode=shared

# Enable Hubble
cilium hubble enable --ui

# Verify
cilium status

Helm Installation

# cilium-values.yaml
kubeProxyReplacement: true

# Ingress controller
ingressController:
  enabled: true
  loadbalancerMode: shared

# L7 proxy (Envoy per node)
envoy:
  enabled: true

# mTLS
encryption:
  enabled: true
  type: wireguard  # or ipsec

# Hubble observability  
hubble:
  enabled: true
  relay:
    enabled: true
  ui:
    enabled: true
  metrics:
    enabled:
      - dns
      - drop
      - tcp
      - flow
      - port-distribution
      - httpV2:exemplars=true;labelsContext=source_ip,source_namespace,destination_ip,destination_namespace

# Gateway API (future of ingress)
gatewayAPI:
  enabled: true
helm repo add cilium https://helm.cilium.io
helm upgrade --install cilium cilium/cilium \
  --namespace kube-system \
  -f cilium-values.yaml

mTLS Encryption

Cilium provides transparent mTLS between pods:

# Enable WireGuard encryption (recommended)
cilium config set encryption-type wireguard

# Or IPsec
cilium config set encryption-type ipsec
cilium config set encryption-ipsec-key-file /path/to/key

Verify encryption:

# Check encryption status
cilium encrypt status

# See encrypted flows
hubble observe --protocol encrypted

Traffic Management

L7 Traffic Policies

# Retry and timeout policies
apiVersion: cilium.io/v2
kind: CiliumEnvoyConfig
metadata:
  name: api-server-config
  namespace: production
spec:
  services:
    - name: api-server
      namespace: production
  resources:
    - "@type": type.googleapis.com/envoy.config.route.v3.RouteConfiguration
      name: api-server-route
      virtual_hosts:
        - name: api-server
          domains: ["*"]
          routes:
            - match:
                prefix: "/"
              route:
                cluster: "production/api-server"
                timeout: 30s
                retry_policy:
                  retry_on: "5xx,reset,connect-failure"
                  num_retries: 3
                  per_try_timeout: 10s

Canary Deployments

apiVersion: cilium.io/v2
kind: CiliumEnvoyConfig
metadata:
  name: canary-routing
  namespace: production
spec:
  services:
    - name: api-server
      namespace: production
    - name: api-server-canary
      namespace: production
  resources:
    - "@type": type.googleapis.com/envoy.config.route.v3.RouteConfiguration
      name: canary-route
      virtual_hosts:
        - name: api
          domains: ["*"]
          routes:
            - match:
                prefix: "/"
                headers:
                  - name: "x-canary"
                    exact_match: "true"
              route:
                cluster: "production/api-server-canary"
            - match:
                prefix: "/"
              route:
                weighted_clusters:
                  clusters:
                    - name: "production/api-server"
                      weight: 90
                    - name: "production/api-server-canary"
                      weight: 10

Rate Limiting

apiVersion: cilium.io/v2
kind: CiliumEnvoyConfig
metadata:
  name: rate-limit
  namespace: production
spec:
  services:
    - name: api-server
      namespace: production
  resources:
    - "@type": type.googleapis.com/envoy.config.filter.http.local_ratelimit.v3.LocalRateLimit
      stat_prefix: http_local_rate_limiter
      token_bucket:
        max_tokens: 100
        tokens_per_fill: 100
        fill_interval: 1s
      filter_enabled:
        runtime_key: local_rate_limit_enabled
        default_value:
          numerator: 100
          denominator: HUNDRED

Ingress with Cilium

Cilium can replace your ingress controller:

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: api-ingress
  annotations:
    ingress.cilium.io/loadbalancer-mode: shared
    ingress.cilium.io/tls-passthrough: "false"
spec:
  ingressClassName: cilium
  tls:
    - hosts:
        - api.company.com
      secretName: api-tls
  rules:
    - host: api.company.com
      http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: api-server
                port:
                  number: 8080
# Gateway class (created by Cilium)
apiVersion: gateway.networking.k8s.io/v1
kind: Gateway
metadata:
  name: production-gateway
  namespace: production
spec:
  gatewayClassName: cilium
  listeners:
    - name: https
      port: 443
      protocol: HTTPS
      hostname: "*.company.com"
      tls:
        mode: Terminate
        certificateRefs:
          - name: wildcard-tls
    - name: http
      port: 80
      protocol: HTTP
      hostname: "*.company.com"
      allowedRoutes:
        kinds:
          - kind: HTTPRoute

---
apiVersion: gateway.networking.k8s.io/v1
kind: HTTPRoute
metadata:
  name: api-route
  namespace: production
spec:
  parentRefs:
    - name: production-gateway
  hostnames:
    - "api.company.com"
  rules:
    - matches:
        - path:
            type: PathPrefix
            value: /v1
      backendRefs:
        - name: api-v1
          port: 8080
          weight: 100
    - matches:
        - path:
            type: PathPrefix
            value: /v2
      backendRefs:
        - name: api-v2
          port: 8080
          weight: 90
        - name: api-v2-canary
          port: 8080
          weight: 10

Observability with Hubble

Hubble provides deep network observability:

# Enable Hubble UI
cilium hubble enable --ui
kubectl port-forward -n kube-system svc/hubble-ui 12000:80

# CLI observability
hubble observe --namespace production

# Filter by verdict
hubble observe --verdict DROPPED

# Filter by HTTP
hubble observe --protocol http --http-status 500

# Export to JSON
hubble observe --output json > flows.json

Prometheus Metrics

# ServiceMonitor for Cilium
apiVersion: monitoring.coreos.com/v1
kind: ServiceMonitor
metadata:
  name: cilium
  namespace: monitoring
spec:
  selector:
    matchLabels:
      app.kubernetes.io/name: cilium-agent
  namespaceSelector:
    matchNames:
      - kube-system
  endpoints:
    - port: prometheus
      interval: 15s

Key Metrics

METRIC                              DESCRIPTION
======                              ===========
cilium_forward_count_total          Packets forwarded
cilium_drop_count_total             Packets dropped (with reason)
hubble_flows_processed_total        L7 flows observed
cilium_policy_verdict_total         Policy decisions
cilium_http_request_duration        HTTP latency

Network Policies (L7)

apiVersion: cilium.io/v2
kind: CiliumNetworkPolicy
metadata:
  name: api-l7-policy
  namespace: production
spec:
  endpointSelector:
    matchLabels:
      app: api-server
  ingress:
    - fromEndpoints:
        - matchLabels:
            app: frontend
      toPorts:
        - ports:
            - port: "8080"
              protocol: TCP
          rules:
            http:
              - method: GET
                path: "/api/v1/public/.*"
              - method: POST
                path: "/api/v1/public/.*"
                headers:
                  - "Content-Type: application/json"
    
    - fromEndpoints:
        - matchLabels:
            app: admin
      toPorts:
        - ports:
            - port: "8080"
              protocol: TCP
          rules:
            http:
              - method: ".*"
                path: "/api/.*"

Migration from Istio

  1. Install Cilium alongside Istio
  2. Migrate namespace by namespace
  3. Remove Istio sidecars
  4. Remove Istio control plane
# Label namespace to disable Istio injection
kubectl label namespace production istio-injection=disabled

# Restart pods to remove sidecars
kubectl rollout restart deployment -n production

# Verify Cilium is handling traffic
hubble observe --namespace production

Troubleshooting

Pods can’t communicate:

cilium connectivity test
cilium status --verbose

L7 policies not working:

# Check Envoy is running
kubectl get pods -n kube-system -l k8s-app=cilium-envoy

# Check policy status
cilium policy get

High latency:

# Check for drops
hubble observe --verdict DROPPED

# Check Envoy metrics
curl -s localhost:9901/stats | grep latency

References

======================================== Cilium + eBPF + Service Mesh

No sidecars. Kernel-level networking.

Found this helpful?

Related Posts

Building a Production-Grade Homelab with K3s, Vault, and FluxCD

How I built a fully GitOps-managed Kubernetes homelab on a single mini PC - from unboxing to production. Proxmox bare metal install, K3s cluster, HashiCorp Vault secrets, full observability, and Cloudflare Tunnel.

OpenTelemetry Changed How I Think About Observability

A practical, opinionated take on OpenTelemetry - why it matters, what it actually solves, and how to instrument across Kubernetes, Lambda, ECS, and EC2 without losing your mind.

Identity Aware Proxy: Zero Trust Access for Internal Applications

Deep dive into Identity Aware Proxies - what they are, how they work, and how to implement them with GCP IAP, Pomerium, and OAuth2-Proxy. Includes Terraform and Kubernetes examples.

Comments