python-environment.md

A major point of Polars on-premises is the system requirements regarding the Python version and Python dependencies, which need to be identical on the client, the scheduler, and all workers. The easiest method to achieve this is having a system-wide Python environment and globally installed packages. We recommend however setting up a virtual environment (uv makes this very easy, including maintaining a given Python version).

The minimal requirement for running polars-on-premises is the polars package.

!!! info "Version pinning" Each release of polars-on-premises is pinned to a single polars release, which can be found in the release announcement. shell export PINNED_VERSION=1.35.2 # for instance

System-wide installation

$ uv pip install --break-system-packages -r requirements.txt polars[cloudpickle]==$PINNED_VERSION
$ ./polars-on-premises service --config-path /etc/polars-cloud/config.toml

Virtual Environment

$ uv venv .venv
$ source .venv/bin/activate
$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$(uv run python -c "import sysconfig; print(sysconfig.get_config_var('LIBDIR'))")
$ uv pip install -r requirements.txt polars[cloudpickle]==$PINNED_VERSION
$ ./polars-on-premises service --config-path /etc/polars-cloud/config.toml