Hey there! Let's break down why you're seeing different CPU core counts between your host machine and the Docker container.

1. Docker on macOS runs in a hidden virtual machine #

Your host is a macOS system (we can tell from the darwin label in your Python output), and Docker Desktop for macOS doesn't run containers directly on your Mac's kernel—it uses a lightweight Linux virtual machine (powered by HyperKit) under the hood.

The CPU core count you see inside the container is actually the number of cores allocated to this virtual machine, not your host's full 4 cores. By default, Docker Desktop usually assigns 2 cores to the VM (you can check or adjust this in Docker Desktop's settings under Resources > CPU).

2. Python's cpu_count() behaves differently across versions #

You're using Python 2.7 on your host and Python 3.6 in the container, and their implementations of multiprocessing.cpu_count() have key differences:

• Python 2.7: It just reads the physical CPU core count of the system it's running on (the host Mac in your case) without accounting for any CPU limits from containerization tools like Docker.
• Python 3.4+: Starting with Python 3.4, cpu_count() takes CPU restrictions set via Linux cgroups (the mechanism Docker uses for resource limiting) into account. If you've set a CPU limit on your container (e.g., using the --cpus=1 flag when starting it), Python 3.x will return that limited number instead of the full core count of the VM or host.

3. Explicit CPU limits in Docker #

If you explicitly restrict the container's CPU resources (using flags like --cpus, --cpu-quota, or --cpu-period), the container will only have access to the specified number of cores, and Python 3.x will reflect this in the cpu_count() output. Even without manual limits, the VM's core allocation (mentioned in point 1) is still the bottleneck here.

• Check your Docker Desktop VM's CPU allocation: Open Docker Desktop > Settings > Resources > CPU to see how many cores are assigned to the underlying VM.
• Run a container with explicit core allocation to verify:

  docker run -it --cpus=4 python
  

  Then inside the container, run import multiprocessing; multiprocessing.cpu_count()—it should return 4 (assuming you've allocated at least 4 cores to the Docker VM first).

内容的提问来源于stack exchange，提问作者kurtgn