kubebn / talos-proxmox-kaas

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Kubernetes As a Service (KAAS) in Proxmox

Introduction

The purpose of this lab to demonstrate capabilities of Talos Linux, Sidero (CAPI), FluxCD & Proxmox Operator, and how they can be used to provision k8s clusters in a true GitOps way.


Built With:

k8s in-cluster tools:

Repository structure

├── kubernetes # App manifests synced via FluxCD & FluxCD configurations
├── manifests # App (system-components) applied via Talos Linux controlplane templates
├── packer # Builds Talos disk on top of Arch Linux for cloud-init functionalities
├── terraform # Proxmox & Talos terraform providers to provision Talos Management cluster
    ├── templates # Management Cluster configurations
    ├── output.tf # Terraform output
    ├── master-nodes.tf # Proxmox master nodes - Management cluster
    ├── worker-nodes.tf # Proxmox worker nodes - Management cluster
    ├── k8s.tf # Fetches Talosconfig & Kubeconfig
    ├── variables.tf # Terraform variables
    ├── terraform.tfvars # Variables to be set here
    ├── talos.tf # Talos provider generates secrets, encodes configuration templates, and applies them to the machines
    └── versions.tf # Terraform providers

Overview

The lab is divided into four stages:

  • Setting up the Proxmox nodes and preparing the cluster, with terraform variables set. This stage is not automated.
  • Building and templating the Talos image using packer. This process can also be done manually, with instructions provided in the documentation.
  • Setting the terraform.tfvars and running terraform to create the Management k8s cluster. The cluster includes:
    • 3 Masters & 3 workers
    • Cilium in strict, tunnel disabled mode automatically synced with the k8s API through the api.cluster.local domain.
    • Talos & Proxmox CCM
    • Metrics-Server
    • CordeDNS-local
    • MetalLB in L2 mode
    • Prometheus CRD's
    • Bootstrapped and installed FluxCD, which syncs the following apps:
      • cert-manager
      • dhcp server
      • proxmox operator and CSI plugin
      • Sidero & CAPI
      • Additionally, it creates cluster-0 demo cluster
  • Sidero cluster bootstrap

Table of contents

(back to top)

Prerequisites

CLI

You will need these CLI tools installed on your workstation

  • talosctl CLI:
curl -sL https://talos.dev/install | sh

Proxmox Node & Cluster configuration

This lab consists of a single Proxmox node with Masquerading (NAT) with iptables configured

auto lo
iface lo inet loopback

auto enp42s0
iface enp42s0 inet static
        address 192.168.1.100/24
        gateway 192.168.1.1

iface wlo1 inet manual

auto vmbr0
iface vmbr0 inet static
        address 10.1.1.1/24
        bridge-ports none
        bridge-stp off
        bridge-fd 0
        post-up   echo 1 > /proc/sys/net/ipv4/ip_forward
        post-up   iptables -t nat -A POSTROUTING -s '10.1.1.0/24' -o enp42s0 -j MASQUERADE
        post-down iptables -t nat -D POSTROUTING -s '10.1.1.0/24' -o enp42s0 -j MASQUERADE
        # wireguard vpn virtual machine config
        post-up   iptables -t nat -A PREROUTING -p udp -d 192.168.1.100 --dport 52890 -i enp42s0 -j DNAT --to-destination 10.1.1.2:52890
        post-down iptables -t nat -A PREROUTING -p udp -d 192.168.1.100 --dport 52890 -i enp42s0 -j DNAT --to-destination 10.1.1.2:52890

The lab infrastructure is provisioned using a flat 10.1.1.0/24 network. Make sure to change the variables according to your networking setup. In addition, ensure that the Proxmox storage names in the terraform.tfvars file are correct.

Although there is only one Proxmox node, a cluster (cluster-1) was initialized using the UI for testing purposes of Proxmox CCM & Talos CCM & Proxmox CSI.

To build the Talos image with cloud-init functionalities, download the Arch Linux image from here, use the following command on the Proxmox machine:

wget -nc -q --show-progress -O "/var/lib/vz/template/iso/archlinux-2023.04.01-x86_64.iso" "http://archlinux.uk.mirror.allworldit.com/archlinux/iso/2023.04.01/archlinux-2023.04.01-x86_64.iso"

You will need to use that image in "iso_file" packer here

(back to top)

Cilium CNI configuration

The following Helm chart template was used to generate a plain yaml manifest, which is then applied in the Talos control plane template:

helm template cilium \
    cilium/cilium \
    --version 1.13.2 \
    --namespace cilium \
    --set ipam.mode=kubernetes \
    --set tunnel=disabled \
    --set bpf.masquerade=true \
    --set endpointRoutes.enabled=true \
    --set kubeProxyReplacement=strict \
    --set autoDirectNodeRoutes=true \
    --set localRedirectPolicy=true \
    --set operator.rollOutPods=true \
    --set rollOutCiliumPods=true \
    --set ipv4NativeRoutingCIDR="10.244.0.0/16" \
    --set hubble.relay.enabled=true \
    --set hubble.ui.enabled=true \
    --set securityContext.capabilities.ciliumAgent="{CHOWN,KILL,NET_ADMIN,NET_RAW,IPC_LOCK,SYS_ADMIN,SYS_RESOURCE,DAC_OVERRIDE,FOWNER,SETGID,SETUID}" \
    --set=securityContext.capabilities.cleanCiliumState="{NET_ADMIN,SYS_ADMIN,SYS_RESOURCE}" \
    --set=cgroup.autoMount.enabled=false \
    --set=cgroup.hostRoot=/sys/fs/cgroup \
    --set=k8sServiceHost="api.cluster.local" \
    --set=k8sServicePort="6443"

Talos clusterDNS & ExtraHostEntries were added to the controlplane & worker configurations:

clusterDNS:
- 169.254.2.53
- ${cidrhost(split(",",serviceSubnets)[0], 10)}

---

- interface: dummy0
addresses:
    - 169.254.2.53/32
extraHostEntries:
- ip: 127.0.0.1
aliases:
    - ${apiDomain} 

By using the configuration from the coredns-local manifest, we can avoid to create separate cilium manifests for each k8s cluster. This is because each cluster has a different k8s cluster API endpoint. To accomplish this, the api.cluster.local domain is added to the coredns configuration, allowing us to apply the same cilium CNI manifest to multiple clusters.

(back to top)

Pull Through Image Cache

In order to speed up the provisioning of the clusters and their components, we use Talos Pull Through Image Cache & Harbor registry. Although this lab does not include the creation and configuration of a Docker registry, automation for Proxmox may be added in the future. If you do not wish to use the Pull image cache, you can delete the following lines in controlplane & workers:

  registries:
    mirrors:
      docker.io:
        endpoints:
          - http://${registry-endpoint}/v2/proxy-docker.io
        overridePath: true
      ghcr.io:
        endpoints:
          - http://${registry-endpoint}/v2/proxy-ghcr.io
        overridePath: true
      gcr.io:
        endpoints:
          - http://${registry-endpoint}/v2/proxy-gcr.io
        overridePath: true
      registry.k8s.io:
        endpoints:
          - http://${registry-endpoint}/v2/proxy-registry.k8s.io
        overridePath: true
      quay.io:
        endpoints:
          - http://${registry-endpoint}/v2/proxy-quay.io
        overridePath: true

(back to top)

Installation

DHCP Disabled

To facilitate the testing of this lab, we need to disable the DHCP service in the infrastructure network; in my configuration vm network is 10.1.1.0/24. This is necessary because a DHCP server will be running inside the Management k8s cluster using FluxCD.

DHCP configuration needs to be changed according to your networking setup - dhcp-config

Variables

The following variables need to be set in terminal:

export PROXMOX_HOST="https://px.host.com:8006/api2/json"
export PROXMOX_TOKEN_ID='root@pam!fire'
export PROXMOX_TOKEN_SECRET="secret"
export PROXMOX_NODE_NAME="proxmox"
export CLUSTER_NAME="mgmt-cluster"

Packer

Configure variables in local.pkrvars.hcl

cd packer

packer init -upgrade .

packer build -only=release.proxmox.talos -var-file="vars/local.pkrvars.hcl" -var proxmox_username="${PROXMOX_TOKEN_ID}" \
-var proxmox_token="${PROXMOX_TOKEN_SECRET}" -var proxmox_nodename="${PROXMOX_NODE_NAME}" -var proxmox_url="${PROXMOX_HOST}" .
Output:
release.proxmox.talos: output will be in this color.

==> release.proxmox.talos: Creating VM
==> release.proxmox.talos: No VM ID given, getting next free from Proxmox
==> release.proxmox.talos: Starting VM
==> release.proxmox.talos: Waiting 25s for boot
==> release.proxmox.talos: Typing the boot command
==> release.proxmox.talos: Using SSH communicator to connect: 10.1.1.30
==> release.proxmox.talos: Waiting for SSH to become available...
==> release.proxmox.talos: Connected to SSH!
==> release.proxmox.talos: Provisioning with shell script: /var/folders/p9/q4lh62654_zbp_psdzkf5gj00000gn/T/packer-shell3807895858
==> release.proxmox.talos:   % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
==> release.proxmox.talos:                                  Dload  Upload   Total   Spent    Left  Speed
==> release.proxmox.talos:   0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
==> release.proxmox.talos:   0     0    0     0    0     0      0      0 --:--:-- --:--:-- --:--:--     0
==> release.proxmox.talos: 100 75.3M  100 75.3M    0     0  9789k      0  0:00:07  0:00:07 --:--:-- 10.6M
==> release.proxmox.talos: 2551808+0 records in
==> release.proxmox.talos: 2551808+0 records out
==> release.proxmox.talos: 1306525696 bytes (1.3 GB, 1.2 GiB) copied, 14.2868 s, 91.5 MB/s
==> release.proxmox.talos: Stopping VM
==> release.proxmox.talos: Converting VM to template
Build 'release.proxmox.talos' finished after 2 minutes 14 seconds.

==> Wait completed after 2 minutes 14 seconds

==> Builds finished. The artifacts of successful builds are:
--> release.proxmox.talos: A template was created: 101


(back to top)

Manual method

We create an Arch Linux VM in Proxmox. Find the name of system disk, for example - local-lvm:vm-106-disk-0, lvm volume vm-106-disk-0. We copy Talos system disk using talos nocloud image to this volume.

cd /tmp
wget https://github.com/siderolabs/talos/releases/download/v1.4.1/nocloud-amd64.raw.xz
xz -d -c nocloud-amd64.raw.xz | dd of=/dev/mapper/vg0-vm--106--disk--0

And then stop & convert that VM into the template in Proxmox.

Terraform

Configure variables in terraform.tfvars

!!! Keep apiDomain variable the same = api.cluster.local. Otherwise, Cilium init is going to fail.

terraform plan

cd terraform

terraform plan -var-file="terraform.tfvars" -var proxmox_token_id="${PROXMOX_TOKEN_ID}" \
 -var proxmox_token_secret="${PROXMOX_TOKEN_SECRET}" -var target_node_name="${PROXMOX_NODE_NAME}" \
 -var proxmox_host="${PROXMOX_HOST}" -var cluster_name="${CLUSTER_NAME}"

terraform apply

terraform apply -auto-approve -var-file="terraform.tfvars" -var proxmox_token_id="${PROXMOX_TOKEN_ID}" \
 -var proxmox_token_secret="${PROXMOX_TOKEN_SECRET}" -var target_node_name="${PROXMOX_NODE_NAME}" \
 -var proxmox_host="${PROXMOX_HOST}" -var cluster_name="${CLUSTER_NAME}"
Output:
# output truncated
local_sensitive_file.talosconfig: Creation complete after 0s [id=542ee0511df16825d846eed4e0bf4f6ca5fdbe61]
null_resource.kubeconfig: Creating...
null_resource.kubeconfig: Provisioning with 'local-exec'...
null_resource.kubeconfig (local-exec): Executing: ["/bin/sh" "-c" "talosctl kubeconfig --force -n 10.1.1.5 -e 10.1.1.5 --talosconfig ./talosconfig"]
null_resource.kubeconfig: Creation complete after 0s [id=5310785605648426604]
null_resource.kubeconfigapi: Creating...
null_resource.kubeconfigapi: Provisioning with 'local-exec'...
null_resource.kubeconfigapi (local-exec): Executing: ["/bin/sh" "-c" "kubectl --kubeconfig ~/.kube/config config set clusters.mgmt-cluster.server https://10.1.1.20:6443"]
null_resource.kubeconfigapi (local-exec): Property "clusters.mgmt-cluster.server" set.
null_resource.kubeconfigapi: Creation complete after 0s [id=3224005877932970184]

Apply complete! Resources: 17 added, 0 changed, 0 destroyed.

Outputs:

cp = <sensitive>
talosconfig = <sensitive>
worker = <sensitive>


After terraform apply is completed, within 10 minutes (depends if you use pull cache or not) you should have the following kubectl output:

kubectl get node -o wide
NAME       STATUS   ROLES           AGE     VERSION   INTERNAL-IP   EXTERNAL-IP   OS-IMAGE         KERNEL-VERSION   CONTAINER-RUNTIME
master-0   Ready    control-plane   2m24s   v1.27.1   10.1.1.5      <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21
master-1   Ready    control-plane   2m30s   v1.27.1   10.1.1.6      <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21
master-2   Ready    control-plane   2m45s   v1.27.1   10.1.1.7      <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21
worker-0   Ready    <none>          2m49s   v1.27.1   10.1.1.8      <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21
worker-1   Ready    <none>          2m33s   v1.27.1   10.1.1.9      <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21
worker-2   Ready    <none>          2m31s   v1.27.1   10.1.1.10     <none>        Talos (v1.4.2)   6.1.27-talos     containerd://1.6.21

kubectl get pod -A
NAMESPACE        NAME                                                READY   STATUS    RESTARTS        AGE
cabpt-system     cabpt-controller-manager-bcbb75fd8-fbczg            1/1     Running   0               70s
cacppt-system    cacppt-controller-manager-5b99d8794f-rphl2          1/1     Running   0               70s
capi-system      capi-controller-manager-86c6bfd9b5-g6xk7            1/1     Running   0               70s
cert-manager     cert-manager-555cc9b8b5-8snqq                       1/1     Running   0               106s
cert-manager     cert-manager-cainjector-55c69fbf8-l77qj             1/1     Running   0               106s
cert-manager     cert-manager-webhook-65ddf78f48-mwq74               1/1     Running   0               106s
cilium           cilium-25ltk                                        1/1     Running   0               2m54s
cilium           cilium-8lhqv                                        1/1     Running   0               2m47s
cilium           cilium-dhhk4                                        1/1     Running   0               2m56s
cilium           cilium-kswnj                                        1/1     Running   0               3m8s
cilium           cilium-m9wfj                                        1/1     Running   0               3m12s
cilium           cilium-operator-7496b89b79-cpxkl                    1/1     Running   0               3m13s
cilium           cilium-operator-7496b89b79-fhfhs                    1/1     Running   0               3m13s
cilium           cilium-qcwgq                                        1/1     Running   0               2m53s
cilium           hubble-relay-84c586cc86-7kpts                       1/1     Running   0               3m13s
cilium           hubble-ui-694cf76f4c-pjnv7                          2/2     Running   0               3m12s
csi-proxmox      proxmox-csi-plugin-controller-666957fd94-rpz97      5/5     Running   0               117s
csi-proxmox      proxmox-csi-plugin-node-5gfng                       3/3     Running   2 (106s ago)    117s
csi-proxmox      proxmox-csi-plugin-node-6vqvs                       3/3     Running   0               117s
csi-proxmox      proxmox-csi-plugin-node-kq2l6                       3/3     Running   2 (105s ago)    117s
flux-system      helm-controller-79ff5d8665-6bnxw                    1/1     Running   0               119s
flux-system      image-automation-controller-679b595d96-62lwq        1/1     Running   0               119s
flux-system      image-reflector-controller-9b7d45fc5-bc7gh          1/1     Running   0               119s
flux-system      kustomize-controller-5b658b9864-9b5rv               1/1     Running   0               119s
flux-system      notification-controller-86d886486b-zb497            1/1     Running   0               119s
flux-system      source-controller-6fd5cb556d-kznjv                  1/1     Running   0               116s
kube-system      coredns-d779cc7ff-mmhj7                             1/1     Running   0               3m10s
kube-system      coredns-d779cc7ff-s2kg8                             1/1     Running   0               3m10s
kube-system      coredns-local-87cfn                                 1/1     Running   0               2m29s
kube-system      coredns-local-bs9q4                                 1/1     Running   0               2m19s
kube-system      coredns-local-dgsp6                                 1/1     Running   0               2m27s
kube-system      coredns-local-jmqc5                                 1/1     Running   0               2m26s
kube-system      coredns-local-pmxp9                                 1/1     Running   0               2m54s
kube-system      coredns-local-qsj7z                                 1/1     Running   0               2m39s
kube-system      dhcp-talos-dhcp-server-7855bb8897-998zn             1/1     Running   0               2m
kube-system      kube-apiserver-master-0                             1/1     Running   0               2m8s
kube-system      kube-apiserver-master-1                             1/1     Running   0               2m23s
kube-system      kube-apiserver-master-2                             1/1     Running   0               2m9s
kube-system      kube-controller-manager-master-0                    1/1     Running   1 (3m32s ago)   2m4s
kube-system      kube-controller-manager-master-1                    1/1     Running   2 (3m36s ago)   2m8s
kube-system      kube-controller-manager-master-2                    1/1     Running   1 (3m17s ago)   2m7s
kube-system      kube-scheduler-master-0                             1/1     Running   1 (3m32s ago)   2m6s
kube-system      kube-scheduler-master-1                             1/1     Running   2 (3m37s ago)   112s
kube-system      kube-scheduler-master-2                             1/1     Running   1 (3m17s ago)   2m9s
kube-system      metrics-server-7b4c4d4bfd-x6dzg                     1/1     Running   0               2m51s
kube-system      proxmox-cloud-controller-manager-79c9ff5cf6-xpn6l   1/1     Running   0               2m49s
kube-system      proxmox-operator-5c79f67c66-w9g72                   1/1     Running   0               117s
kube-system      talos-cloud-controller-manager-776fdbd456-wtmjk     1/1     Running   0               2m46s
metallb-system   controller-7948676b95-w58sw                         1/1     Running   0               2m55s
metallb-system   speaker-9h28g                                       1/1     Running   0               2m10s
metallb-system   speaker-dkrvr                                       1/1     Running   0               2m11s
metallb-system   speaker-drxxr                                       1/1     Running   0               2m11s
metallb-system   speaker-fz8ls                                       1/1     Running   0               2m10s
metallb-system   speaker-j2v4g                                       1/1     Running   0               2m10s
metallb-system   speaker-nbpl7                                       1/1     Running   0               2m10s
sidero-system    caps-controller-manager-5b4b95bcdb-74bvb            1/1     Running   0               70s
sidero-system    sidero-controller-manager-7d5b96cf6d-tppq6          4/4     Running   0               70s

terraform output for talosconfig & kubeconfig can be checked if needed

terraform output -raw talosconfig
terraform output -raw kubeconfig

CSI testing

Storageclasses should be already configured in the cluster:

kubectl get sc
NAME               PROVISIONER                RECLAIMPOLICY   VOLUMEBINDINGMODE      ALLOWVOLUMEEXPANSION   AGE
proxmox-data       csi.proxmox.sinextra.dev   Delete          WaitForFirstConsumer   true                   6m33s
proxmox-data-xfs   csi.proxmox.sinextra.dev   Delete          WaitForFirstConsumer   true                   6m33s

Let's apply a test Pod with volume attached and check on that:

kubectl apply -f https://raw.githubusercontent.com/sergelogvinov/proxmox-csi-plugin/main/docs/deploy/test-pod-ephemeral.yaml

kubectl apply -f https://raw.githubusercontent.com/sergelogvinov/proxmox-csi-plugin/main/docs/deploy/test-pod-ephemeral.yaml
pod/test created

kubectl -n default get pods,pvc
NAME       READY   STATUS    RESTARTS   AGE
pod/test   1/1     Running   0          70s

NAME                             STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS       AGE
persistentvolumeclaim/test-pvc   Bound    pvc-41b1aea6-fa99-4ad3-b28d-57f1ce4a85aa   1Gi        RWO            proxmox-data-xfs   70s

---
kubectl describe pv pvc-41b1aea6-fa99-4ad3-b28d-57f1ce4a85aa
Name:              pvc-41b1aea6-fa99-4ad3-b28d-57f1ce4a85aa
Labels:            <none>
Annotations:       pv.kubernetes.io/provisioned-by: csi.proxmox.sinextra.dev
                   volume.kubernetes.io/provisioner-deletion-secret-name:
                   volume.kubernetes.io/provisioner-deletion-secret-namespace:
Finalizers:        [kubernetes.io/pv-protection external-attacher/csi-proxmox-sinextra-dev]
StorageClass:      proxmox-data-xfs
Status:            Bound
Claim:             default/test-pvc
Reclaim Policy:    Delete
Access Modes:      RWO
VolumeMode:        Filesystem
Capacity:          1Gi
Node Affinity:
  Required Terms:
    Term 0:        topology.kubernetes.io/region in [cluster-1]
                   topology.kubernetes.io/zone in [proxmox]
Message:
Source:
    Type:              CSI (a Container Storage Interface (CSI) volume source)
    Driver:            csi.proxmox.sinextra.dev
    FSType:            xfs
    VolumeHandle:      cluster-1/proxmox/vms/vm-9999-pvc-41b1aea6-fa99-4ad3-b28d-57f1ce4a85aa
    ReadOnly:          false
    VolumeAttributes:      storage=vms
                           storage.kubernetes.io/csiProvisionerIdentity=1683735975451-8081-csi.proxmox.sinextra.dev
Events:                <none>

Try StatefulSet:

kubectl apply -f https://raw.githubusercontent.com/sergelogvinov/proxmox-csi-plugin/main/docs/deploy/test-statefulset.yaml

kubectl -n default get pods,pvc -owide                                                                                                                                                                                                                                                          4s ⎈ admin@mgmt-cluster
NAME         READY   STATUS    RESTARTS   AGE   IP             NODE       NOMINATED NODE   READINESS GATES
pod/test-0   1/1     Running   0          57s   10.244.3.95    worker-2   <none>           <none>
pod/test-1   1/1     Running   0          57s   10.244.0.119   worker-0   <none>           <none>
pod/test-2   1/1     Running   0          57s   10.244.2.234   worker-1   <none>           <none>

NAME                                   STATUS    VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE   VOLUMEMODE
persistentvolumeclaim/storage-test-0   Bound     pvc-d8be278f-ffee-49c3-b303-10fc2ebd79ae   1Gi        RWO            proxmox-data   57s   Filesystem
persistentvolumeclaim/storage-test-1   Bound     pvc-a7b73455-1ae2-4fb5-a7ca-1a671e81491c   1Gi        RWO            proxmox-data   57s   Filesystem
persistentvolumeclaim/storage-test-2   Bound     pvc-f7ab10a6-50af-40d4-b704-8fefb0d1bff9   1Gi        RWO            proxmox-data   57s   Filesystem

(back to top)

Sidero Bootstrap

If you check your Proxmox UI, you might find 2 new VM's already running in there. What we can see in the Management k8s cluster though:

kubectl get qemu

NAME              STATUS   POWER   CLUSTER     NODE      VMID
sidero-master-1   SYNCED   ON      cluster-1   proxmox   108
sidero-worker-1   SYNCED   ON      cluster-1   proxmox   109
---

kubectl get servers

NAME                                   HOSTNAME    ACCEPTED   CORDONED   ALLOCATED   CLEAN   POWER   AGE
f0ac3f32-ee63-11ed-a05b-0242ac120003   10.1.1.52   true                              true    on      12m
fe4fadea-ee63-11ed-a05b-0242ac120003   10.1.1.53   true                              true    on      12m
---

kubectl get serverclasses

AME               AVAILABLE                                                                         IN USE   AGE
any                ["f0ac3f32-ee63-11ed-a05b-0242ac120003","fe4fadea-ee63-11ed-a05b-0242ac120003"]   []       14m
master-cluster-0   []                                                                                []       14m
worker-cluster-0   []                                                                                []       14m
---

kubectl get cluster 

NAME        PHASE         AGE   VERSION
cluster-0   Provisioned   14m
---

kubectl get TalosControlPlane

NAME           READY   INITIALIZED   REPLICAS   READY REPLICAS   UNAVAILABLE REPLICAS
cluster-0-cp                         1                           1
---

kubectl get MachineDeployment

NAME                CLUSTER     REPLICAS   READY   UPDATED   UNAVAILABLE   PHASE       AGE   VERSION
cluster-0-workers   cluster-0   1                  1         1             ScalingUp   14m   v1.27.1

As mentioned before, all of these manifests were applied and synced via FluxCD from cluster-0.

If you look at ServerClasses for masters and workers, labelSelectors are specified in there hence we need to apply labels to the servers in order to start bootstrapping a new k8s cluster.

kubectl label servers fe4fadea-ee63-11ed-a05b-0242ac120003 worker-dev=true
kubectl label servers f0ac3f32-ee63-11ed-a05b-0242ac120003 master-dev=true

server.metal.sidero.dev/fe4fadea-ee63-11ed-a05b-0242ac120003 labeled
server.metal.sidero.dev/f0ac3f32-ee63-11ed-a05b-0242ac120003 labeled

At this time, we can see that both servers are now In Use, which means that cluster creation is initiliazed:

kubectl get serverclasses

NAME               AVAILABLE   IN USE                                                                            AGE
any                []          ["f0ac3f32-ee63-11ed-a05b-0242ac120003","fe4fadea-ee63-11ed-a05b-0242ac120003"]   22m
master-cluster-0   []          ["f0ac3f32-ee63-11ed-a05b-0242ac120003"]                                          22m
worker-cluster-0   []          ["fe4fadea-ee63-11ed-a05b-0242ac120003"]                                          22m

Now, lets fetch talosconfig and kubeconfig from the cluster:

kubectl get talosconfig -o yaml $(kubectl get talosconfig --no-headers | awk 'NR==1{print $1}') -o jsonpath='{.status.talosConfig}' > cluster-0.yaml

talosctl --talosconfig cluster-0.yaml kubeconfig --force -n 10.1.1.40 -e 10.1.1.40

Due to the fact that we use Cilium for all of our clusters with api.cluster.local, we want to change the API endpoint to the Talos cluster VIP, which in my terraform settings is set to 10.1.1.40:

kubectl --kubeconfig ~/.kube/config config set clusters.cluster-0.server https://10.1.1.40:6443

Property "clusters.cluster-0.server" set.

At that point, we have a fully working cluster bootstrapped and provisioned via Sidero and FluxCD:

kubectl get node

NAME            STATUS   ROLES           AGE     VERSION
talos-5o3-l4b   Ready    <none>          3m2s    v1.27.1
talos-svd-m63   Ready    control-plane   2m56s   v1.27.1
---

kubectl get pod -A

NAMESPACE        NAME                                    READY   STATUS    RESTARTS        AGE
cilium           cilium-c878h                            1/1     Running   0               3m7s
cilium           cilium-operator-7496b89b79-7zx66        1/1     Running   0               3m13s
cilium           cilium-operator-7496b89b79-ff4vt        1/1     Running   0               3m13s
cilium           cilium-rhqxv                            1/1     Running   0               3m1s
cilium           hubble-relay-84c586cc86-z82jt           1/1     Running   0               3m13s
cilium           hubble-ui-694cf76f4c-bzlq2              2/2     Running   0               3m13s
kube-system      coredns-5665966b56-7zz8m                1/1     Running   0               3m13s
kube-system      coredns-5665966b56-vbtwq                1/1     Running   0               3m13s
kube-system      coredns-local-7pls2                     1/1     Running   0               2m36s
kube-system      coredns-local-qlkcg                     1/1     Running   0               2m34s
kube-system      kube-apiserver-talos-svd-m63            1/1     Running   0               2m39s
kube-system      kube-controller-manager-talos-svd-m63   1/1     Running   2 (3m44s ago)   2m17s
kube-system      kube-scheduler-talos-svd-m63            1/1     Running   2 (3m44s ago)   2m1s
kube-system      kubelet-csr-approver-7759f94756-bkhzr   1/1     Running   0               3m13s
kube-system      metrics-server-7b4c4d4bfd-h2sc6         1/1     Running   0               3m8s
metallb-system   controller-7948676b95-947g5             1/1     Running   0               3m12s
metallb-system   speaker-6kzj8                           1/1     Running   0               2m34s
metallb-system   speaker-m7d7z                           1/1     Running   0               2m36s

Scaling

In case if you want to add more nodes, you can add QEMU objects in vms.yaml manifest. Even though we have all resources in the management cluster, we can't scale the nodes using kubectl command like here thats because we use FluxCD to sync our manifests in the cluster. Therefore, to add more worker nodes we need to scale replicas in cluster-0.yaml manifest MachineDeployment object. Additionally, all of these new nodes need to be labeled with worker-dev=true labelSelectors.

Likewise, if you add more nodes via Proxmox-operator, you can either set your custom uuid's for the nodes or delete the whole smbios1 part.

smbios1: "uuid=f0ac3f32-ee63-11ed-a05b-0242ac120003,manufacturer=MTIz,product=MTIz,version=MTIz,serial=MTIz,sku=MTIz,family=MTIz,base64=1"

New clusters

If you need to create a new cluster, follow this doc. It will be the same process of running these commands, and once cluster-1 manifest is generated you can add it to FluxCD git repo.

export CONTROL_PLANE_SERVERCLASS=master-cluster-1
export WORKER_SERVERCLASS=worker-cluster-1
export TALOS_VERSION=v1.3.0
export KUBERNETES_VERSION=v1.27.1
export CONTROL_PLANE_PORT=6443
export CONTROL_PLANE_ENDPOINT=api.cluster.local

clusterctl generate cluster cluster-1 -i sidero:v0.5.8 > cluster-1.yaml



(back to top)

Teraform destroy

terraform destroy -refresh=false -auto-approve -var-file="terraform.tfvars" -var proxmox_token_id="${PROXMOX_TOKEN_ID}" \
 -var proxmox_token_secret="${PROXMOX_TOKEN_SECRET}" -var target_node_name="${PROXMOX_NODE_NAME}" -var proxmox_host="${PROXMOX_HOST}" -var cluster_name="${CLUSTER_NAME}"

Bear in mind that if you run terraform destroy, it's not going to delete VM's which were provisioned by Proxmox Operator/FluxCD. You would need to stop sync in FluxCD, and destroy those machines via kubectl delete qemu vm-name command.

References

https://www.sidero.dev/v0.5/getting-started/

https://www.talos.dev/v1.4/talos-guides/install/cloud-platforms/nocloud/

https://github.com/sergelogvinov/terraform-talos

Author's blog

https://bnovickovs.me/

terraform gitignore template

# Local .terraform directories
**/.terraform/*

# .tfstate files
*.tfstate
*.tfstate.*

# Crash log files
crash.log
crash.*.log

# Ignore override files as they are usually used to override resources locally and so
# are not checked in
override.tf
override.tf.json
*_override.tf
*_override.tf.json

# Include override files you do wish to add to version control using negated pattern
# !example_override.tf

# Include tfplan files to ignore the plan output of command: terraform plan -out=tfplan
# example: *tfplan*

# Ignore CLI configuration files
.terraformrc
terraform.rc
.terraform.lock.hcl

talosconfig
kubeconfig

About


Languages

Language:HCL 64.2%Language:Smarty 35.8%