- Docs
- Networking
Networking
Expose applications
- Choose how to expose an application Decide between a LoadBalancer Service, an ingress controller, or a NodePort before you make a workload reachable from the internet.
- Expose an app with a LoadBalancer Service Place a single application on the internet with a type LoadBalancer Service that the Hetzner cloud-controller-manager turns into a load balancer.
- Proxy without a load balancer Point DNS straight at your node IPs and run Traefik or nginx there, with hostPort on the pod network so the host firewall needs no policy.
- Open a port on the nodes Allow a port in the default-deny host firewall with a CiliumClusterwideNetworkPolicy, so a hostNetwork workload is reachable from outside the cluster.
Load balancing
- Configure a Hetzner load balancer Tune the control-plane load balancer with topology variables and per-Service load balancers with annotations, including the full annotation reference and bare-metal targets.
- Tune health checks for node replacement Set load balancer health checks so a node being drained and replaced leaves the target pool before it drops traffic.
- Preserve the client source IP Stop your pods from seeing the load balancer address instead of the real client. Set up PROXY protocol, use externalTrafficPolicy Local, and handle hairpinning.
- Serve IPv6 clients Accept IPv6 connections through the Hetzner load balancer while every node and pod stays IPv4-only.
- Pass TLS through at Layer 4 Terminate TLS inside your pod instead of at the ingress by passing raw TCP or TLS straight through the load balancer.
Ingress and Gateway API
- Install Traefik for HTTP routing Place many HTTP applications behind one Hetzner load balancer with Traefik. The platform includes no ingress controller, so you install the one you prefer.
- Route traffic with the Gateway API Use the Gateway API (Gateway and HTTPRoute) instead of Ingress for role-split routing and richer traffic rules on Syself Autopilot.
- Serve multiple domains and wildcards Host many client domains and wildcard subdomains behind one ingress and one Hetzner load balancer.
Certificates
- Terminate TLS at the ingress Hold the certificate at the ingress controller, decrypt HTTPS there, and forward plain HTTP to your pods.
- Install cert-manager Install cert-manager as a normal workload so the cluster can request and renew TLS certificates on its own.
- Set up a Let's Encrypt ClusterIssuer Create a Let's Encrypt ClusterIssuer and let cert-manager fetch free certificates for every ingress in the cluster.
- Rotate and renew certificates How cert-manager renews certificates before they expire, and how to force, monitor, and recover a rotation.
Egress and IP addressing
- Egress gateway for a stable source IP Route outbound traffic from chosen pods through one node so an IP-allowlisted third party always sees the same source address.
- Configure pod and service subnets Set the pod and service CIDRs before you create a cluster, because they are fixed at creation and cannot change afterward.
- MTU and the VXLAN tunnel How Cilium's VXLAN overlay wraps pod traffic and how the resulting MTU affects throughput between nodes.
- Connect to on-prem or another VPC Reach a private network, on-prem site, or another cloud VPC from the cluster over a VPN, planning IP ranges so nothing overlaps.
Service mesh and resilience
- Choose a service mesh Decide whether you need a service mesh at all, and which one, before you add sidecars to every pod.
- Handle unreliable networks with Istio Use Istio retries and outlier detection so brief packet loss between nodes turns into a slightly slower response instead of a visible error.
DNS
- Understand in-cluster DNS (CoreDNS) How CoreDNS resolves Service names, how to resolve private names through dnsServers, the ndots cost, and how Syself autoscales CoreDNS.
- Automate DNS records with external-dns Let external-dns publish and update A and AAAA records for your Services and Ingresses automatically at your DNS provider.
Debugging and troubleshooting
- Debug node networking Diagnose the most common node-level network failures using the files and service logs that show what a node actually sees.
- Diagnose pod DNS resolution Track down why a pod cannot resolve a name, from CoreDNS health to a policy that blocks port 53.
- Run a Cilium connectivity test Use Cilium's built-in connectivity test and Hubble to prove pod-to-pod, pod-to-Service, and egress paths actually work.
- Troubleshoot load balancers and service traffic A Service has no external IP or traffic is dropped, so trace it from the Hetzner CCM through the host firewall to the pod.