Ingress controller setup on HA Kubernetes Cluster¶
This guide provides instruction how to setup experimental ingress controller feature. For this, we are hosting our cluster on OpenStack VMs and using the Rancher Kubernetes Engine (RKE) to deploy and manage our Kubernetes Cluster and ingress controller
The result at the end of this tutorial will be:
- Customization of the cluster.yaml file for ingress controller support
- Installation and configuration test DNS server for ingress host resolution on testing machines
- Instalation and configuration MLB (Metal Load Balancer) required for exposing ingress service
- Instalation and configuration NGINX ingress controller
- Additional info howto deploy onap with services exposed via Ingress controller
Before setup cluster for ingress purposes DNS cluster IP and ingress provider should be configured and follwing:
Where the <cluster_dns_ip> should be set to the same IP as the CONTROLPANE node.
For external load balacer purposes minimum one of the worker node should be configured with external IP address accessible outside the cluster. It can be done using the following example node configuration:
Where the <external_ip> is external worker node IP address, and <internal_ip> is internal node IP address if it is required
DNS server deployed on the Kubernetes cluster makes it easy to use services exposed through ingress controller because it resolves all subdomain related to the onap cluster to the load balancer IP. Testing ONAP cluster requires a lot of entries on the target machines in the /etc/hosts. Adding many entries into the configuration files on testing machines is quite problematic and error prone. The better wait is to create central DNS server with entries for all virtual host pointed to simpledemo.onap.org and add custom DNS server as a target DNS server for testing machines and/or as external DNS for kubernetes cluster.
DNS server has automatic instalation and configuration script, so instalation is quite easy:
> cd kubernetes/contrib/dns-server-for-vhost-ingress-testing > ./deploy\_dns.sh
After DNS deploy you need to setup DNS entry on the target testing machine. Because DNS listen on non standard port configuration require iptables rules on the target machine. Please follow the configuation proposed by the deploy scripts Example output depends on the IP address and example output looks like bellow:
DNS server already deployed: 1. You can add the DNS server to the target machine using following commands: sudo iptables -t nat -A OUTPUT -p tcp -d 192.168.211.211 --dport 53 -j DNAT --to-destination 10.10.13.14:31555 sudo iptables -t nat -A OUTPUT -p udp -d 192.168.211.211 --dport 53 -j DNAT --to-destination 10.10.13.14:31555 sudo sysctl -w net.ipv4.conf.all.route_localnet=1 sudo sysctl -w net.ipv4.ip_forward=1 2. Update /etc/resolv.conf file with nameserver 192.168.211.211 entry on your target machine
By default pure Kubernetes cluster requires external load balancer if we want to expose external port using LoadBalancer settings. For this purpose MetalLB can be used. Before installing the MetalLB you need to ensure that at least one worker has assigned IP acessible outside the cluster.
MetalLB Load balanancer can be easily installed using automatic install script:
> cd kubernetes/contrib/metallb-loadbalancer-inst > ./install-metallb-on-cluster.sh
After installation DNS server and ingress controller we can install and configure ingress controller. It can be done using the following commands:
> cd kubernetes/contrib/ingress-nginx-post-inst > kubectl apply -f nginx_ingress_cluster_config.yaml > kubectl apply -f nginx_ingress_enable_optional_load_balacer_service.yaml
After deploy NGINX ingress controller you can ensure that the ingress port is exposed as load balancer service with external IP address:
> kubectl get svc -n ingress-nginx NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE default-http-backend ClusterIP 10.10.10.10 <none> 80/TCP 25h ingress-nginx LoadBalancer 10.10.10.11 10.12.13.14 80:31308/TCP,443:30314/TCP 24h