Skip to content
This repository has been archived by the owner on Nov 22, 2023. It is now read-only.

frank-at-suse/vsphere_cilium_kubeproxy-less

Repository files navigation

RKE2 Cluster with kube-proxy -less Cilium CNI & BBR Pod Congestion Control

Rancher Terraform Kubernetes

Reason for Being

This Terraform plan is for creating a kube-proxy -less multi-node RKE2 cluster using Cilium CNI. The Cilium chart enables BBR Pod Congestion Control for greatly increased network performance as well. Also installed is Rancher's Prometheus Operator Chart (cluster_monitoring.tf) along with an HAProxy Ingress Grafana Dashboard to gain visibility into your Ingress traffic (this plan disables the built-in RKE2 NGiNX Ingress in favor of HAProxy because...reasons ¯\_(ツ)_/¯ ). kube-vip makes a token appearance here to advertise the HAProxy Ingress Controller via ARP.

This is a nice boiler-plate Terraform plan for a high-performing cluster that includes a very capable monitoring stack.

Environment Prerequisites

  • Functional Rancher Management Server with vSphere Cloud Credential

  • vCenter >= 7.x and credentials with appropriate permissions (see https://github.com/rancher/barn/blob/main/Walkthroughs/vSphere/Permissions/README.md)

  • Virtual Machine Hardware Compatibility at Version >= 15

  • Linux Kernel >= 5.18 (required for enabling BBR Pod Congestion Control)

  • Create the following in the files/ directory:

    NAME PURPOSE
    .rancher-api-url URL for Rancher Management Server
    .rancher-bearer-token API bearer token generated via Rancher UI
    .ssh-public-key SSH public key for additional OS user

Caveats

  • Cilium's Hubble UI is disabled as it can be a drag on performance. However, if you enjoy looking at groupings of rectangles connected with lines and do want to enable Hubble, reference the RKE2 Cilium Helm chart HERE.

  • If you don't want to run the Rancher Prometheus Operator, it can be uninstalled at any time simply by removing cluster_monitoring.tf from your working directory and re-applying the plan. It is here as a demonstration/value-add, not a requirement of any kind.

  • kube-vip is operating via ARP mode, so services published via LoadBalancer will have traffic directed to a single node.

  • Unlike RKE2's "baked-in" NGiNX Ingress Controller, HAProxy's ingress is not FIPS 140-2 compliant.

To Run

terraform apply

Tested Versions

SOFTWARE VERSION DOCS
kube-vip 0.6.2 https://kube-vip.io/docs
Rancher Prometheus Operator 102.0.2+up40.1.2 https://docs.ranchermanager.rancher.io/pages-for-subheaders/monitoring-and-alerting
Rancher Server 2.7.6 https://rancher.com/docs/rancher/v2.6/en/overview
Rancher Terraform Provider 3.1.1 https://registry.terraform.io/providers/rancher/rancher2/latest/docs
RKE2 1.26.8+rke2r1 https://docs.rke2.io
Terraform 1.4.6 https://www.terraform.io/docs
vSphere 8.0.1.00300 https://docs.vmware.com/en/VMware-vSphere/index.html

About

Terraform plan for creating a multi-node kube-proxy -less RKE2 cluster w/ Cilium CNI on VMware vSphere

Topics

Resources

License

Stars

Watchers

Forks

Languages