Kubernetes Pod Placement Basics: A Complete Guide to nodeName and nodeSelector ☸️

When operating a Kubernetes cluster, situations often arise where you need to explicitly place a specific pod on a desired node. Today, we will provide a detailed overview of nodeName and nodeSelector, the most intuitive and fundamental scheduling methods.

---

1. What is Pod Scheduling? 📍

In Kubernetes, scheduling is the process of determining which appropriate node a pending pod will be assigned to and executed on. By default, the kube-scheduler automatically places pods by checking the node’s resource status, but when an administrator needs direct control, nodeName and nodeSelector are used.

---

2. The Most Powerful and Simple Method: nodeName 🎯

nodeName completely bypasses the scheduler’s complex calculations, forcing a pod to be placed on a specific node by explicitly naming it.

• Features: It bypasses the kube-scheduler and directly requests the kubelet on the specified node to run the pod.
• Advantages: Very fast and definitive.
• Disadvantages: Absolutely no flexibility. If the node with that name does not exist or lacks resources, the pod will not run.

💻 Example YAML Configuration

YAML

apiVersion: v1
kind: Pod
metadata:
  name: nodename-pod
spec:
  containers:
  - name: nginx
    image: nginx
  nodeName: worker-node-01 # Must be assigned to this node only

---

3. Flexible Selection Based on Labels: nodeSelector 🏷️

nodeSelector is a method that selects a group of nodes for pod assignment based on the

labels

set on the nodes. It is currently the most recommended basic scheduling method in Kubernetes.

• Features: Pods are placed only on nodes that have labels matching the key-value pairs defined in the pod specification.
• Usage Flow:

1. Assign labels to nodes.
2. Specify the corresponding labels in the nodeSelector of the pod’s YAML.

🛠️ Assigning Labels to Nodes

First, assign unique characteristics to nodes from the terminal.

Bash

kubectl label nodes worker-node-02 disktype=ssd

💻 Example YAML Configuration

YAML

apiVersion: v1
kind: Pod
metadata:
  name: nodeselector-pod
spec:
  containers:
  - name: nginx
    image: nginx
  nodeSelector:
    disktype: ssd # Placed only on nodes with the 'disktype=ssd' label

---

4. nodeName vs nodeSelector Comparison Summary 📊

Category	nodeName	nodeSelector
—	—	—
Method	Directly specify node name	Specify node labels as conditions
Scheduler Intervention	X (Bypasses scheduler)	O (Scheduler searches for conditions)
Flexibility	Very low (Fixed)	Medium (Selects among nodes with same labels)
Recommended Use	For debugging specific node failures	When specific hardware (GPU, SSD) is required

—

5. Cautions and Limitations ⚠️

1. Case Sensitivity: Label keys and values are strictly case-sensitive.
2. Match Requirement: All labels specified in nodeSelector must exist on the node. If even one does not match, the pod will remain in a Pending state.
3. Complex Conditions: Complex logic such as “Node A or B”, or “exclude nodes without a label” is difficult to implement with nodeSelector. In such cases, the higher-level concept of Node Affinity should be used.

We hope this information helps you efficiently manage your cluster resources.