LearnGlossary

What are resource limits in Kubernetes?

Updated 25 October 2026 · Published by Zoorik

A CPU limit in Kubernetes is the most CPU time a container may use. A Kubernetes memory limit caps its memory.

Kubernetes resource limits: CPU and memory

On Linux, the kernel enforces each one differently:

  • A container at its CPU limit is slowed, not killed. This is CPU throttling.
  • Over its memory limit, a container may get an out of memory (OOM) kill.
  • With no limit or default, a container can use any free CPU or memory on its node.
  • Set a limit with no request, and Kubernetes copies the limit into the request, unless a default request applies.

An example from the Kubernetes docs

The docs’ frontend pod has two containers. Each requests 250m of CPU and 64 MiB of memory, with limits of 500m and 128 MiB. 250m means 250 millicpu, a quarter of one CPU.

At its CPU limit a container slows. Over its memory limit it may be OOM killed. Kubernetes docs example: each of the 2 containers of the frontend pod. Two bars, each drawn from 0 to its own limit, with the request filled in. CPU: request 250m, halfway along the bar, and the bar ends at a line labelled limit 500m, slowed here, not killed. Memory: request 64 MiB, halfway along the bar, and the bar ends at a line labelled limit 128 MiB, above it the container may be OOM killed. At its CPU limit a container slows. Over its memory limit it may be OOM killed. Kubernetes docs example: each container of the frontend podCPUrequest 250mlimit 500mslowed here, not killedMemoryrequest 64 MiBlimit 128 MiBabove it: may be OOM killed
The numbers from the Kubernetes docs: each bar runs from 0 to its own limit, and the filled part is the request, 250m of CPU and 64 MiB of memory.

Why it matters for your bill

The scheduler places pods by their requests, not their limits. But when a limit becomes the request, tight limits can add nodes:

  • A tight CPU limit throttles the app.
  • A tight memory limit can lead to OOM kills.
  • Raising the limit to fix this raises the request, so fewer pods fit per node.

Our guide to requests and limits covers both.

Coral rebalances CPU and memory requests on AKS and EKS to match real use. On regular pods it never reduces a CPU or memory limit. If a raised memory request needs it, the limit rises to 1.5 times the request.

With requests that match use, the same pods fit on fewer nodes. The free assessment works out your own figure.

Read next: Kubernetes pod rebalancing

Sources: Kubernetes, “Resource Management for Pods and Containers” (opens in a new tab); Kubernetes, “Assign CPU Resources to Containers and Pods” (opens in a new tab).

Find your own number

One line in your cloud console. One PDF report with your estimated saving at list price.

Estimate your saving

Talk to the builders

Thirty minutes with the people who build Coral.

Book a demo (opens in a new tab)

Or write to founders@zoorik.com