Skip to content
Deadline to Die

Module 2: Scheduling & Placement

🏗️ 1. The Manual Task: Capacity Planning

Without a scheduler, an SRE must manually decide where each application runs.

The Manual Steps:

  1. Check Resources: SSH into node-1 and check free memory (free -m).
  2. Assign Workload: Decide if the node has enough RAM.
  3. Run App: Pull the Docker image and run the container manually.
  4. Update Logs: Update a central inventory system (like an Excel sheet).

📍 2. The Kubernetes Abstraction: kube-scheduler

The kube-scheduler automates this capacity planning using a two-step process:

  • Filtering (Predicates): It filters out nodes that can’t run the Pod (e.g., node is out of memory).
  • Scoring (Priorities): It ranks the remaining nodes (e.g., prefer nodes where the image is already pulled).
  • NodeSelector: Allows you to steer workloads based on node labels (e.g., “only run on nodes with GPUs”).