argocd-autopilot VS helmfile

# get the nodes in the cluster data "proxmox_virtual_environment_nodes" "proxmox_nodes" {} # VM Definition resource "proxmox_virtual_environment_vm" "example" { count = var.vm_count name = count.index + 1 <= var.vm_masters ? "${var.vm_name}-master-${format("%02d", count.index + 1)}" : "${var.vm_name}-worker-${format("%02d", count.index - (var.vm_masters - 1))}" node_name = data.proxmox_virtual_environment_nodes.proxmox_nodes.names[count.index % length(data.proxmox_virtual_environment_nodes.proxmox_nodes.names)] vm_id = count.index + 1 <= var.vm_masters ? var.vm_proxmox_id + count.index : var.vm_proxmox_id + count.index + (var.vm_proxmox_id_offset - var.vm_masters) tags = sort(concat(var.vm_proxmox_tags, [count.index + 1 <= var.vm_masters ? "master" : "worker"] )) agent { enabled = true trim = true } cpu { sockets = var.vm_sockets cores = var.vm_cores } memory { dedicated = count.index + 1 <= var.vm_masters ? var.vm_mem_master : var.vm_mem_worker } disk { interface = "scsi0" datastore_id = var.clone_target_local ? var.clone_target_datastore_local : var.clone_target_datastore_nfs ssd = true size = count.index + 1 <= var.vm_masters ? var.vm_disk_size_master : var.vm_disk_size_worker iothread = true discard = "on" } network_device { model = "virtio" mac_address = count.index + 1 <= var.vm_masters ? "${var.net_mac_address_base}AA:${format("%02d", count.index)}" : "${var.net_mac_address_base}BB:${format("%02d", count.index - var.vm_masters)}" # vlan_id = var.net_vlan_id # Not needed since using dedicated interface bridge = var.net_bridge } serial_device {} # clone information clone { vm_id = var.clone_target_local ? var.clone_vm_id + (count.index % var.vm_masters) : var.clone_vm_id datastore_id = var.clone_target_local ? var.clone_target_datastore_local : var.clone_target_datastore_nfs node_name = var.clone_target_local ? data.proxmox_virtual_environment_nodes.proxmox_nodes.names[count.index % length(data.proxmox_virtual_environment_nodes.proxmox_nodes.names)]: data.proxmox_virtual_environment_nodes.proxmox_nodes.names[0] } # had to add a wait for agent to come alive provisioner "remote-exec" { inline = [ "sudo cloud-init status --wait", "sudo systemctl start qemu-guest-agent", ] connection { type = "ssh" agent = false port = 22 host = element(element(self.ipv4_addresses, index(self.network_interface_names, "eth0")), 0) private_key = file(var.public_key_path) user = var.vm_username } } } # Create file for ansible inventory resource "local_file" "k3s_file" { content = templatefile( "${path.module}/templates/inventory_ansible.tftpl", { ansible_masters = "${join("\n", [for vm in slice(proxmox_virtual_environment_vm.example, 0, var.vm_masters) : join("", [vm.ipv4_addresses[1][0] ])])}" ansible_nodes = "${join("\n", [for vm in slice(proxmox_virtual_environment_vm.example, var.vm_masters , var.vm_count) : join("", [vm.ipv4_addresses[1][0] ])])}" } ) filename = "${path.module}/../ansible-k3s/inventory/k3s-cluster/hosts.ini" } #connecting to the Ansible control node and call ansible playbook to build the k3s cluster resource "null_resource" "call-ansible" { provisioner "local-exec" { command = "ANSIBLE_HOST_KEY_CHECKING=False ansible-playbook ${path.module}/../ansible-k3s/site.yml -i ${path.module}/../ansible-k3s/inventory/k3s-cluster/hosts.ini" } depends_on = [ local_file.k3s_file ] } #Copy the kubectl file locally so we can issue commands against the cluster resource "null_resource" "copy-kubeconfig" { provisioner "local-exec" { command = "scp -o 'StrictHostKeyChecking no' seb@${proxmox_virtual_environment_vm.example[0].ipv4_addresses[1][0]}:~/.kube/config ~/.kube/config " } depends_on = [ null_resource.call-ansible ] } #bootstrap the cluster with argocd-autopilot resource "null_resource" "argocd-autopilot" { provisioner "local-exec" { command = ( var.first_install ? "argocd-autopilot repo bootstrap --repo ${var.github_repo} -t ${var.github_token} --app https://github.com/argoproj-labs/argocd-autopilot/manifests/ha" : "argocd-autopilot repo bootstrap --recover --app ${var.github_repo}.git/bootstrap/argo-cd" ) } depends_on = [ null_resource.copy-kubeconfig ] }

I use Argo CD Autopilot which bootstraps Argo CD in a self-managing structure. If nothing else, copy the repo structure https://github.com/argoproj-labs/argocd-autopilot

We use the same approach internally and we fully open-sourced our solution at https://argocd-autopilot.readthedocs.io/en/stable/

I recommend utilising Autopilot a companion project that not only installs Argo CD but also commits all configurations to git so Argo CD can maintain itself using GitOps.

Check https://argocd-autopilot.readthedocs.io/en/stable/ It is an installer that does exactly that. It installs ArgoCD, sets it up to manage itself and offers a suggested bootstrap for your applications

Checkout the ArgoCD autopilot if you're using kustomize rather than helm

sudo wget https://github.com/helmfile/helmfile/releases/download/v0.159.0/helmfile_0.159.0_linux_amd64.tar.gz sudo tar -xxf helmfile_0.159.0_linux_amd64.tar.gz sudo rm helmfile_0.159.0_linux_amd64.tar.gz sudo mv helmfile /usr/local/bin/

Installation: https://github.com/helmfile/helmfile/releases

What are the benefits over using helmfile? https://helmfile.readthedocs.io/

Helm charts and Helmfile

And now I can just run dudo -l "https://github.com/helmfile/helmfile/releases/download/v{{ version }}/helmfile_{{ version }}_{{ os }}_{{ arch }}.tar.gz" -i /tmp/helmfile.tar.gz -p $HELMFILE_VERSION

I might be misunderstanding your question but we use https://github.com/helmfile/helmfile along with Argo, so essentially between eks and those I could rebuild our entire cluster in minutes.

I am asking which do you choose, Flux or Helmfile. edit: and what criteria do you use to select.

do you publish your own helm chart for your internal services and use it in every environment? if so, you could try to use helmfile within the service's repo itself and store values in a helm/$env directory. then enhance your ci to deploy to dev after the merge/image build phase directly. to try and cut out what sounds like a "deployment/config repo" step you have in the middle that's making everything a pain.

I've used helmfile before to declaratively manage multiple helm charts. It's a higher-level tool, and still uses helm under the hood.