Summary of "Day 13/40 - Static Pods, Manual Scheduling, Labels, and Selectors in Kubernetes"

Summary of “Day 13/40 - Static Pods, Manual Scheduling, Labels, and Selectors in Kubernetes”

This video is part of a Kubernetes tutorial series focused on advanced scheduling concepts and metadata management within Kubernetes clusters. The main topics covered include:

1. Kubernetes Scheduler and Pod Scheduling Process

The scheduler is a control plane component responsible for deciding which node a pod should run on.
When a pod creation request is sent via kubectl to the API server, the scheduler monitors pods in a pending state and assigns them to appropriate nodes based on scheduling algorithms.
The scheduler then instructs the kubelet on the chosen node to start the pod.
The scheduler itself runs as a static pod on the control plane node.

2. Static Pods

Static pods are pods managed directly by the kubelet, not by the scheduler.
They are typically used for running control plane components like the API server, controller manager, scheduler, etc.
Static pod manifests are stored in a specific directory (/etc/kubernetes/manifests) on the control plane node.
The kubelet continuously monitors this directory and ensures these pods are running.
Removing a static pod manifest (e.g., the scheduler manifest) stops that component; this can be used for troubleshooting or restarting control plane components.

3. Manual Scheduling

Manual scheduling involves specifying the nodeName field in a pod’s spec to assign it to a specific node.
Pods with nodeName set are not scheduled by the scheduler; kubelet on the specified node runs them directly.
This method allows pods to be scheduled even if the scheduler is down.
Demonstrated by creating a pod YAML with nodeName set and showing it runs without the scheduler.

4. Labels and Selectors

Labels are key-value pairs attached to Kubernetes objects (pods, deployments, services, etc.) to organize and select subsets of resources.
Labels help filter and group resources, similar to tags or filters on e-commerce websites.
Selectors are used to query or filter resources based on labels (e.g., select pods with tier=frontend).
Labels are applied in the metadata section of objects; selectors are used in commands like kubectl get pods --selector=tier=frontend.
Labels can be added or modified dynamically on running pods.
Annotations differ from labels; they store metadata about the object (e.g., last applied configuration) but are not used for selection.
Difference between namespaces and labels:
- Namespaces provide logical separation and security boundaries for resources.
- Labels are flexible tags for grouping and filtering resources within or across namespaces.

5. Hands-On Demonstrations and Assignments

The video includes live demos showing:

How to inspect static pod manifests on the control plane node.
How to stop/start the scheduler by moving static pod manifests.
Manual scheduling by specifying nodeName.
Adding labels to pods and querying with selectors.

An assignment is provided for viewers to practice these concepts on their home labs. Viewers are encouraged to engage via comments and a Discord channel for peer support.

Main Speakers / Source

The video is presented by P, the channel owner and instructor of the CK 2024 Kubernetes series.

Key Takeaways

Understanding static pods is crucial for managing Kubernetes control plane components.
Manual scheduling provides a fallback mechanism when the scheduler is unavailable.
Labels and selectors are powerful tools for organizing and managing Kubernetes resources efficiently.
Practical knowledge of these concepts is essential for Kubernetes certification and real-world cluster management.

This video is a valuable resource for learners preparing for the Certified Kubernetes Administrator (CKA) exam and those seeking deeper insight into Kubernetes scheduling and resource organization.