Deep Dive: SIG Scheduling
summary
Deep dive into Kubernetes scheduler internals with concrete features and progress: explains how the scheduler tracks node capacity, the priority and preemption behavior (preemption shipped around v1.10, race conditions possible because decisions use snapshots), how taints and tolerations affect placement, and the implemented extension for pod group gang scheduling (annotation-based pod group + controller with min/desired members, supports strict and best‑effort models). Notes which pieces are implemented, which are prototype/discussed (reservation of total resources and lifecycle restart per-member are planned but not finished).
AI-written summary · about
chapters (4)
related talks
35mSIG Scheduling Deep Dive
45mSIG-Scheduling Intro & Deep Dive
25mSIG Scheduling Intro and Deep Dive
30mDeep Dive: Scheduling SIG
39mIntro: Scheduling SIG
34mKubernetes Scheduling Features or How Can I Make the System Do What I Want?
39mSIG-Scheduling Deep Dive
42mDeep Dive Into the Latest Kubernetes Scheduler Features