.. _mlflow-troubleshooting:

Troubleshooting
===============

MLflow is not installed
   The integration layer can exist without the upstream package. Install a
   compatible MLflow distribution in the runtime environment where tracking is
   required.

CLI flag rejected
   Inspect ``get_mlflow_server_cli_caps`` rather than hard-coding flags from a
   different MLflow release.

Server starts but readiness fails
   Check bind host/port, backend store, artifact root, process stderr and proxy or
   container networking. Process creation alone is not readiness.

Project root is wrong
   Inspect marker configuration and the ``start`` path. Keep project discovery
   deterministic rather than depending on the shell's current directory.

Security policy
   See :ref:`mlflow-security-and-readiness` for the current enforcement caveat.