Kubernetes
Practical
By Samson Tanimawo, PhD
Published Dec 2, 2025
4 min read
Cluster Resource Allocation
Where does cluster capacity go? Audit.
Live workflow · 3 working · 1 queuedLive
Signal · gather Working
Decide · pick action Working
Apply · with verify Working
Learn · update playbook Queued
Buckets
System overhead, kube-system, app workloads, idle.
Each measurable.
Typical
System: 10-20%. App: 50-70%. Idle: 15-30%.
Idle = waste.
Act
Reduce idle. VPA, HPA tuning.
Compounds.