kubeadm join - Node Cluster Membership

Overview

kubeadm join adds nodes to an existing cluster through bidirectional trust establishment.

Core principle: Join requires two-way trust - the node must verify the control plane (Discovery), and the control plane must verify the node (TLS Bootstrap).

When to Use

• Adding worker nodes to cluster
• Adding control plane nodes for HA
• Troubleshooting node join failures
• Understanding TLS bootstrap mechanism

Bidirectional Trust

┌─────────────────────────────────────────────────────────────┐
│                    BIDIRECTIONAL TRUST                      │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  Discovery (Node → Control Plane)                          │
│  ─────────────────────────────────                         │
│  "Is this API Server legitimate?"                          │
│  Method: CA cert hash verification or kubeconfig file      │
│                                                             │
│  TLS Bootstrap (Control Plane → Node)                      │
│  ────────────────────────────────────                      │
│  "Is this node authorized to join?"                        │
│  Method: Bootstrap token → CSR → Certificate issuance      │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Basic Join Command

From kubeadm init output:

kubeadm join 192.168.10.100:6443 \
  --token abcdef.0123456789abcdef \
  --discovery-token-ca-cert-hash sha256:xxxx...

Worker Node Join

Phases

1. preflight     → Check requirements
2. discovery     → Get cluster-info, verify CA
3. TLS bootstrap → Authenticate with token, submit CSR, get certificate
4. kubelet-start → Configure and start kubelet

Command

kubeadm join 192.168.10.100:6443 \
  --token 123456.1234567890123456 \
  --discovery-token-ca-cert-hash sha256:bd763182...

Post-Join Files (Worker)

/etc/kubernetes/
├── kubelet.conf          # kubelet kubeconfig
└── pki/
    └── ca.crt            # Cluster CA (for verification)

/var/lib/kubelet/
├── config.yaml           # kubelet configuration
└── kubeadm-flags.env     # Extra kubelet flags

Control Plane Node Join (HA)

Additional Phases

1. preflight
2. discovery
3. control-plane-prepare  → Download certs, generate manifests
4. TLS bootstrap
5. etcd-join             → Add to etcd cluster
6. kubelet-start
7. control-plane-join    → Apply labels/taints

Command

kubeadm join 192.168.10.100:6443 \
  --token 123456.1234567890123456 \
  --discovery-token-ca-cert-hash sha256:bd763182... \
  --control-plane \
  --certificate-key <certificate-key>

Get Certificate Key

# On existing control plane
kubeadm init phase upload-certs --upload-certs
# Outputs: certificate-key: <64-char-hex>

# Or during initial kubeadm init
kubeadm init --upload-certs

Token Management

List Tokens

kubeadm token list

Create New Token

# Simple
kubeadm token create

# With join command
kubeadm token create --print-join-command

# With custom TTL
kubeadm token create --ttl 2h

Get CA Hash

openssl x509 -pubkey -in /etc/kubernetes/pki/ca.crt | \
  openssl rsa -pubin -outform der 2>/dev/null | \
  openssl dgst -sha256 -hex | sed 's/^.* //'

Discovery Methods

Token-Based (Default)

kubeadm join ... --token xxx --discovery-token-ca-cert-hash sha256:xxx

• Fetches cluster-info ConfigMap from kube-public namespace
• Verifies CA with provided hash

File-Based

kubeadm join ... --discovery-file /path/to/cluster-info.yaml

• Uses pre-distributed kubeconfig file
• For air-gapped or highly secure environments

Unsafe (Not Recommended)

kubeadm join ... --discovery-token-unsafe-skip-ca-verification

• Skips CA verification
• Vulnerable to MITM attacks

TLS Bootstrap Flow

New Node                           Control Plane
────────                           ─────────────
    │                                    │
    │──── 1. Get cluster-info ──────────▶│ (unauthenticated)
    │◀─── CA cert + API endpoint ────────│
    │                                    │
    │──── 2. Authenticate with token ───▶│
    │◀─── Temporary credentials ─────────│
    │                                    │
    │──── 3. Submit CSR ────────────────▶│
    │                                    │ (auto-approved)
    │◀─── 4. Signed certificate ─────────│
    │                                    │
    │──── 5. Use certificate ───────────▶│ (full kubelet access)
    │                                    │

Configuration File Approach

# join-config.yaml
apiVersion: kubeadm.k8s.io/v1beta4
kind: JoinConfiguration
discovery:
  bootstrapToken:
    apiServerEndpoint: "192.168.10.100:6443"
    token: "123456.1234567890123456"
    caCertHashes:
      - "sha256:bd763182..."
nodeRegistration:
  kubeletExtraArgs:
    - name: node-ip
      value: "192.168.10.101"
  criSocket: "unix:///run/containerd/containerd.sock"

kubeadm join --config=join-config.yaml

Common Issues

Note: The error "could not find a JWS signature in the cluster-info ConfigMap for token ID" means your bootstrap token has expired. The JWS signature is removed from cluster-info when the token expires.

Troubleshooting Steps

1. Verify API Server Reachability

# From worker node
curl -k https://192.168.10.100:6443/healthz
# Expected: "ok"

2. Check Token Validity

# On control plane
kubeadm token list
# Check expiration time

3. Verify Network

# From worker
ping 192.168.10.100
ss -tlnp | grep 6443  # Should show kubelet after join

4. Check Firewall

# On control plane
systemctl is-active firewalld
# If active, open required ports or disable
firewall-cmd --permanent --add-port=6443/tcp
firewall-cmd --permanent --add-port=10250/tcp
firewall-cmd --reload

5. Check kubelet Logs

journalctl -u kubelet -f

Reset and Retry

# On the node that failed to join
kubeadm reset -f
rm -rf /etc/kubernetes /var/lib/kubelet
iptables -F && iptables -t nat -F

# Re-attempt join
kubeadm join ...

Verification Checklist

After successful join:

• Node appears in kubectl get nodes
• Node becomes Ready (after CNI installed)
• kubelet service running: systemctl status kubelet
• kube-proxy running: crictl ps | grep proxy
• /etc/kubernetes/kubelet.conf exists
• /var/lib/kubelet/config.yaml exists

cluster-info ConfigMap

The bootstrap process depends on this publicly-accessible ConfigMap:

# View (works without authentication)
curl -k https://192.168.10.100:6443/api/v1/namespaces/kube-public/configmaps/cluster-info

# Contains:
# - kubeconfig: API Server address + CA certificate
# - jws-kubeconfig-<token>: Signed verification

This is the ONLY Kubernetes resource accessible without authentication.