Skip to content

Latest commit

 

History

History
116 lines (78 loc) · 4.49 KB

README.md

File metadata and controls

116 lines (78 loc) · 4.49 KB
Raw

Installation | Configuration | Resources Displayed | Contributing

jupyter-resource-usage

Github Actions Status Binder PyPI conda-forge PyPI GitHub

Screenshot with memory limit

Jupyter Resource Usage is an extension for Jupyter Notebooks and JupyterLab that displays an indication of how much resources your current notebook server and its children (kernels, terminals, etc) are using. This is displayed in the main toolbar in the notebook itself, refreshing every 5s.

Kernel resource usage can be displayed in a sidebar for IPython kernels with ipykernel >= 6.11.0.

Installation

You can currently install this package from PyPI.

pip install jupyter-resource-usage

Or with conda:

conda install -c conda-forge jupyter-resource-usage

If your notebook version is < 5.3, you need to enable the extension manually.

jupyter serverextension enable --py jupyter_resource_usage --sys-prefix
jupyter nbextension install --py jupyter_resource_usage --sys-prefix
jupyter nbextension enable --py jupyter_resource_usage --sys-prefix

Configuration

Memory Limit

jupyter-resource-usage can display a memory limit (but not enforce it). You can set this in several ways:

  1. MEM_LIMIT environment variable. This is set by JupyterHub if using a spawner that supports it.
  2. In the commandline when starting jupyter notebook, as --ResourceUseDisplay.mem_limit.
  3. In your Jupyter notebook traitlets config file

The limit needs to be set as an integer in Bytes.

Memory usage warning threshold

Screenshot with memory warning

The background of the resource display can be changed to red when the user is near a memory limit. The threshold for this warning can be configured as a fraction of the memory limit.

If you want to flash the warning to the user when they are within 10% of the memory limit, you can set the parameter --ResourceUseDisplay.mem_warning_threshold=0.1.

CPU Usage

jupyter-resource-usage can also track CPU usage and report a cpu_percent value as part of the /api/metrics/v1 response.

You can set the cpu_limit in several ways:

  1. CPU_LIMIT environment variable. This is set by JupyterHub if using a spawner that supports it.
  2. In the command line when starting jupyter notebook, as --ResourceUseDisplay.cpu_limit.
  3. In your Jupyter notebook traitlets config file

The limit corresponds to the number of cpus the user has access to, but does not enforce it.

Additionally, you can set the track_cpu_percent trait to enable CPU usage tracking (disabled by default):

c = get_config()
c.ResourceUseDisplay.track_cpu_percent = True

As a command line argument:

jupyter notebook --ResourceUseDisplay.track_cpu_percent=True

Disable Prometheus Metrics

There is a known bug with Prometheus metrics which causes "lag"/pauses in the UI. To workaround this you can disable Prometheus metric reporting using:

--ResourceUseDisplay.enable_prometheus_metrics=False

Resources Displayed

Currently the server extension only reports memory usage (just RSS) and CPU usage. Other metrics will be added in the future as needed.

The notebook extension currently doesn't show CPU usage, only memory usage.

Contributing

If you would like to contribute to the project, please read the CONTRIBUTING.md. The CONTRIBUTING.md file explains how to set up a development installation and how to run the test suite.