CPU throttling
Pausing a container for part of each time slice because it reached its CPU limit, which makes it slow rather than killing it.
In more detail
The kernel enforces the limit through control groups, so a container short of CPU keeps running but responds late and may fail probes. Unlike memory, hitting a CPU limit never produces OOMKilled. Throttling statistics in the cgroup and latency metrics confirm it.
Related terms
- CPU limit — A cap on how much processor time a container may use, expressed as a number of processors.
- control group — A Linux kernel feature that limits how much memory, processor time and other resources a group of processes may use.
- OOM killer — The part of the Linux kernel that kills a process when memory runs out.
- resource limit — A cap on how much memory, processor time or number of processes a container may use.
Taught in
- Requests, Limits, Quotas And Horizontal Autoscaling — Kubernetes: Production-Grade Container Orchestration
- Troubleshooting: Pending Pod Triage — Kubernetes: Production-Grade Container Orchestration
- Troubleshooting: OOMKilled, Throttling And Evictions — Kubernetes: Production-Grade Container Orchestration