Module 4
Scheduling
The scheduler picks a node for each new Pod in two phases, filter then score. You steer it with requests, taints, affinity, spread rules and priorities.
Why it matters. Most Pending Pods, noisy-neighbour problems and surprise outages after a node failure are scheduling decisions. Reading the scheduler's reasoning is a core operations skill.
After this module you can answer
- How does the scheduler choose between 200 nodes?
- Why do requests drive placement but limits drive enforcement?
- Which Pod is evicted first under memory pressure?
- How do I keep replicas out of one failure domain?
- When can one Pod evict another?
What is inside
1Filter and score
2Framework
3Requests, limits, QoS
4Quota and LimitRange
5Taints and tolerations
6Node affinity
7Spread across zones
8Priority and PDB
9Module infographic