Configure Pulse to Monitor MLflow

You can configure Pulse to monitor MLflow performance and enable observability by updating the Python executable path.

Steps

Verify the Python executable used by MLflow

1. On the node where MLflow is running, run the following command:

systemctl status mlflow

Example output:

2. Identify the Python executable path used by MLflow. The example shows that MLflow uses the following Python executable:

/usr/odp/3.6.2-1/mlflow/bin/python3.11

Edit the Configuration File

In the configuration file, update the Python executable path (custom bin) used by MLFlow under the base section.

In Docker-Based Pulse Deployment

• Open the override.yml file in a text editor and update the path.
• File Path: $AcceloHome/work/<cluster_name/override.yml

In Kubernetes-Based Deployment

• In the Admin UI:
  • Click a configured cluster, and go to the Configuration tab.
  • In Configuration, update the path in VARs YAML.
  • For details, see Manage Configuration Files.

base:

  extra_adsvcproc_process_name: "<path-to-custom-python-bin-used-by-mlflow>, <path-to-other-custom-bins>"

Example:

base:

  extra_adsvcproc_process_name: "/usr/odp/3.6.2-1/mlflow/bin/python3.11, <path-to-other-custom-bins>"

Apply the Configuration

In Docker-Based Pulse Deployment

After saving the file, run the following command to apply the changes.

accelo reconfig cluster

In Kubernetes-Based Deployment

After making the update, click Reconfigure to apply the changes in Admin UI.