K8s setup with Kubeadm & ArgoCD

kubernetes/kubernetes#110177

sudo mkdir /etc/containerd
sudo vim /etc/containerd/config.toml

[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc.options]
            BinaryName = ""
            CriuImagePath = ""
            CriuPath = ""
            CriuWorkPath = ""
            IoGid = 0
            IoUid = 0
            NoNewKeyring = false
            NoPivotRoot = false
            Root = ""
            ShimCgroup = ""
            SystemdCgroup = true

• https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/install-kubeadm/
  • We use containerd, need to enable CRI (Docker installation disables this, delete /etc/containerd/config.toml file)
  • Install kubeadm, kubectl, kubelet
• https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/create-cluster-kubeadm/

TODO: UPDATE IP ADDRESS!!!

Grep repo for IP, exists in couple of config files.

sudo kubeadm init \
--skip-phases=addon/kube-proxy \
--apiserver-advertise-address="192.168.86.73" \
--apiserver-cert-extra-sans="kube.aarnihalinen.fi,87.92.163.107"

We are running single-node cluster => need to untaint the control-plane

kubectl taint nodes --all node-role.kubernetes.io/control-plane-

TODO: New setup with proxmox with 2 nodes bootstrapped with talos https://docs.siderolabs.com/talos/v1.11/platform-specific-installations/virtualized-platforms/proxmox

Update README, figure out a way to commit talos configuration to repo

Install argocd CLI

brew install argocd

Install ArgoCD with Kustomize

kubectl kustomize argocd | kubectl apply -f -
kubectl apply -f apps/templates/argo-projects.yml
kubectl apply -f app-of-apps.yml

Port forward to the ArgoCD server

kubectl port-forward svc/argocd-server -n argocd 8080:443

Open ArgoCD UI with

argocd admin dashboard -n argocd

and login (set new password as well)

argocd admin initial-password -n argocd
argocd login localhost:8080
argocd account update-password

Argo should install all other apps (they are included in the app-of-apps).

Cilium and cert-manager are by default multi-node => scale down deployments to 1. Ingress ports might need to be edited to service, seems like a bug.