By Guanru Huang
In the cloud, the high-availability of K8S clusters is ensured by cloud providers. When one node crashes, the cloud provider automatically replaces the node and load balancers redirects the network traffic to other pod replicas in the cluster, without the end-user noticing downtime.
For on-prem k8s clusters, high-availability of the k8s control plane and services is a major challenge. To solve high-availability, Kube-vip can be used to assign virtual IPs for control plane and Metallb can be used to create load-balancers on-prem.
In this talk, I will demonstrate how to build a three-node on-prem k8s HA cluster using Kube-vip and Metallb, how to use Kong proxy as loadbalancers, and why HA is important in the 5G scenario.
Guan-ru is a cloud DevOps engineer at Atayalan, a startup providing universal connectivity platform for industry 4.0. He previously worked as a tech consultant in hybrid cloud at HPE, and wrote his master thesis on cloud migration at SAP in Germany. His technical interests include hybrid cloud infrastructure, automation tooling and security.