Kubernetes.K3s.installLog

3 VM's provisioned with Ubuntu Server 22.04

additional lvm configuration

pvdisplay
pvcreate /dev/sdb
vgdisplay
vgcreate longhorn-vg /dev/sdb
lvdisplay
lvcreate -l 100%FREE -n longhorn-lv longhorn-vg
ls /dev/mapper
mkfs.ext4 /dev/mapper/longhorn--vg-longhorn--lv
#! add "UUID=<uuid> /mnt/blockstorage ext4 defaults 0 0" to /etc/fstab
mkdir /mnt/blockstorage
mount -a

K3s cluster

On first node (replace <floating ip> with the correct value):

curl -sfL https://get.k3s.io | sh -s - server --cluster-init --disable local-storage,servicelb --tls-san <floating ip>
cat /var/lib/rancher/k3s/server/token
kubectl config view --raw

Install kube-vip (replace <interface name> and <floating ip> with the correct values):

ctr image pull ghcr.io/kube-vip/kube-vip:latest
cat << EOF > /var/lib/rancher/k3s/server/manifests/kube-vip.yml
$(curl https://kube-vip.io/manifests/rbac.yaml)
---
$(ctr run --rm --net-host ghcr.io/kube-vip/kube-vip:latest vip /kube-vip manifest daemonset --interface <interface name> --address <floating ip> --inCluster --taint --controlplane --services --arp --leaderElection)
EOF

On subsequent nodes (replace <floating ip> and <value from master> with the correct values):

curl -sfL https://get.k3s.io | K3S_URL=https://<floating ip>:6443 K3S_TOKEN=<value from master> sh -s - server --disable local-storage,servicelb

0) Configure automatic updates

Install Rancher's System Upgrade Controller:

kubectl apply -f https://github.com/rancher/system-upgrade-controller/releases/latest/download/system-upgrade-controller.yaml

Apply a server (master node) and agent (worker node) plan:

kubectl apply -f system/UpgradeController/plan-Server.yml # -f system/UpgradeController/plan-Agent.yml

1) Secret management

Prereq: latest kubeseal release

1.1) Install Helm Chart

See Bitnami Sealed Secrets:

helm repo add sealed-secrets https://bitnami-labs.github.io/sealed-secrets
helm repo update
helm install sealed-secrets-controller -n kube-system sealed-secrets/sealed-secrets

Retrieve public/private keys (store these on a secure location!):

kubectl get secret -n kube-system -l sealedsecrets.bitnami.com/sealed-secrets-key -o yaml > BitnamiSealedSecrets.masterkey.yml

2) Persistent storage

2.1) storageClass for SMB (CIFS):

See https://github.com/kubernetes-csi/csi-driver-smb:

curl -skSL https://raw.githubusercontent.com/kubernetes-csi/csi-driver-smb/master/deploy/install-driver.sh | bash -s master --

Store credentials in secret:

kubectl apply -f storage/csi-driver-smb/sealedSecret-CSIdriverSMB.yml

2.2) flexVolume for SMB (CIFS):

curl -Ls https://github.com/juliohm1978/kubernetes-cifs-volumedriver/blob/master/install.yaml -o storage/flexVolSMB/daemonSet-flexVolSMB.yml

Override drivername to something more sensible (see storage/flexVolSMB/daemonSet-flexVolSMB.yml)

spec:
  template:
    spec:
      containers:
        - image: juliohm/kubernetes-cifs-volumedriver-installer:2.0
          ...
          env:
            - name: VENDOR
              value: mount
            - name: DRIVER
              value: smb
          ...

Perform installation:

kubectl apply -f storage/flexVolSMB/daemonSet-flexVolSMB.yml

Wait for installation to complete (check logs of all installer-pods), then pause daemonSet:

kubectl patch daemonset juliohm-cifs-volumedriver-installer -p '{"spec": {"template": {"spec": {"nodeSelector": {"intentionally-paused": ""}}}}}'

Store credentials in secret:

kubectl apply -f storage/flexVolSMB/sealedSecret-flexVolSMB.yml

2.3) storageClass for distributed block storage:

See Longhorn Helm Chart:

helm repo add longhorn https://charts.longhorn.io && helm repo update
helm install longhorn longhorn/longhorn --namespace longhorn-system --create-namespace --values=storage/Longhorn/chart-values.yml

Log on to the web interface and delete the default disks on each node (mounted at /var/lib/longhorn) and replace them with new disks mounted at /mnt/blockstorage.

Add additional storageClass with backup schedule: After specifying a NFS backup target (syntax: nfs://servername:/path/to/share) through Longhorn's dashboard

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: longhorn-dailybackup
provisioner: driver.longhorn.io
allowVolumeExpansion: true
parameters:
  numberOfReplicas: "3"
  staleReplicaTimeout: "2880"
  fromBackup: ""
  recurringJobs: '[{"name":"backup", "task":"backup", "cron":"0 0 * * *", "retain":14}]'

Then make this the new default storageClass:

kubectl patch storageclass longhorn-dailybackup -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'
#kubectl delete storageclass longhorn

3) Ingress Controller

Reconfigure default Traefik configuration: See Traefik 2.x Helm Chart and HelmChartConfig

kubectl apply -f ingress/Traefik2.x/helmchartconfig-traefik.yaml

4) GitOps

4.1) Install Helm Chart

See ArgoCD:

helm repo add argo https://argoproj.github.io/argo-helm
helm repo update
helm install argo-cd -n argo-cd --create-namespace argo/argo-cd --values system/ArgoCD/chart-values.yml

Retrieve initial password:

kubectl get secret -n argocd argocd-initial-admin-secret -o jsonpath='{.data.password}' | base64 -d; echo

Login with username admin and the initial password, browse to User Info and Update Password.

Create ArgoCD applicationset

kubectl apply -f system/ArgoCD/applicationset-homelab.yml

5) Services

5.1) Argus (release management)

kubectl apply -f services/Argus

5.2) Authelia (single sign-on))

kubectl apply -f services/Authelia

5.3) Vaultwarden (password manager)

Requires mount.cifs' option nobrl

kubectl apply -f services/Vaultwarden

5.4) DDclient (dynamic dns)

kubectl apply -f services/DDclient

5.5) Gitea (git repository)

kubectl apply -f services/Gitea

5.6) Gotify (notifications)

kubectl apply -f services/Gotify

5.7) Guacamole (remote desktop gateway)

Requires specifying a uid & gid in both the securityContext of the db container and the persistentVolume

kubectl apply -f services/Guacamole

Wait for the included containers to start, then perform the following commands to initialize the database:

kubectl exec -n guacamole -i guacamole-<pod-id> --container guacamole -- /opt/guacamole/bin/initdb.sh --postgresql > initdb.sql
kubectl exec -n guacamole -i guacamole-<pod-id> --container db -- psql -Uguacamole -f - < initdb.sql
kubectl rollout restart deployment -n guacamole guacamole

5.8) Lighttpd (webserver)

Serves various semi-containerized websites; respective webcontent is stored on fileshare

kubectl apply -f services/Lighttpd/configMap-Lighttpd.yml
kubectl apply -f services/Lighttpd/deploy-Lighttpd.yml

5.9) PVR namespace (automated media management)

Containers use shared resources to be able to interact with downloaded files

kubectl create secret generic --type=mount/smb smb-secret --from-literal=username=<<omitted>> --from-literal=password=<<omitted>> -n pvr
kubectl apply -f services/PVR/persistentVolumeClaim-PVR.yml
kubectl apply -f services/PVR/storageClass-PVR.yml

5.9.1) Plex (media library)

Due to usage of symlinks, partially incompatible with SMB-share-backed storage

kubectl apply -f services/PVR/deploy-Plex.yml

After deploying, Plex server needs to be claimed (=assigned to Plex-account):

kubectl get endpoints Plex -n PVR

Browse to the respective IP address (http://:32440/web) and follow instructions.

5.9.2) Prowlarr (indexer management)

kubectl apply -f services/PVR/deploy-Prowlarr.yml

5.9.3) Radarr (movie management)

kubectl apply -f services/PVR/deploy-Radarr.yml

5.9.4) SABnzbd (download client)

kubectl apply -f services/PVR/deploy-SABnzbd.yml

5.9.5) Sonarr (tv management)

kubectl apply -f services/PVR/deploy-Sonarr.yml

6) Miscellaneous

Various notes/useful links

• Replacement for not-yet-deprecated kubectl get all -A:

  kubectl get $(kubectl api-resources --verbs=list -o name | paste -sd, -) --ignore-not-found --all-namespaces
  

• DaemonSet to configure nodes' sysctl fs.inotify.max-user-watches:

  kubectl apply -f system/InotifyMaxWatchers/daemonSet-InotifyMaxWatchers.yml
  

• Debug DNS lookups within the cluster:

  kubectl run -it --rm dnsutils --restart=Never --image=gcr.io/kubernetes-e2e-test-images/dnsutils -- nslookup [-debug] [fqdn]
  

  or

  kubectl run -it --rm busybox --restart=Never --image=busybox:1.28 -- nslookup api.github.com [-debug] [fqdn]
  

• Delete namespaces stuck in Terminating state: First check whether there are any resources still present; preventing the namespace from being deleted:

  kubectl api-resources --verbs=list --namespaced -o name \
    | xargs -n 1 kubectl get --show-kind --ignore-not-found -n <namespace>
  

  Any resources returned should be deleted first (worth mentioning: if you get an error error: unable to retrieve the complete list of server APIs, you should check kubectl get apiservice for any apiservice with a status of False) If there are no resources left in the namespace, and it is still stuck terminating, the following commands remove the blocking finalizer (this is a last resort, you are bypassing protections put in place to prevent zombie processes):

  kubectl get namespace <namespace> -o json | jq -j '.spec.finalizers=null' > tmp.json
  kubectl replace --raw "/api/v1/namespaces/<namespace>/finalize" -f ./tmp.json
  rm ./tmp.json