Databricks Workflows

Deploying to Databricks is simplified using the RTDIP SDK as this method of deployment will handle the setup of the libraries and spark configuration directly from the components being used in your pipeline.

Prerequisites

This deployment method expects to deploy a local file to Databricks Workflows

Import

from rtdip_sdk.pipelines.deploy import DatabricksSDKDeploy, CreateJob, JobCluster, ClusterSpec, Task, NotebookTask, ComputeSpecKind, AutoScale, RuntimeEngine, DataSecurityMode

Authentication

Azure Active DirectoryDatabricks

Refer to the Azure Active Directory documentation for further options to perform Azure AD authentication, such as Service Principal authentication using certificates or secrets. Below is an example of performing default authentication that retrieves a token for Azure Databricks.

Also refer to the Code Reference for further technical information.

from rtdip_sdk.authentication import authenticate as auth

authentication = auth.DefaultAuth().authenticate()
access_token = authentication.get_token("2ff814a6-3304-4ab8-85cb-cd0e6f879c1d/.default").token

Note

If you are experiencing any trouble authenticating please see Troubleshooting - Authentication

Refer to the Databricks documentation for further information about generating a Databricks PAT Token. Below is an example of performing default authentication that retrieves a token for a Databricks Workspace.

Provide your dbapi..... token to the access_token in the examples below.

access_token = "dbapi.........."

Deploy

Deployments to Databricks are done using the Databricks SDK. The Databricks SDK enables users to control exactly how they deploy their RTDIP Pipelines to Databricks.

Any of the Classes below can be imported from the following location:

from rtdip_sdk.pipelines.deploy import *

Parameters for a Databricks Job can be managed using the following Classes:

Class	Description
ClusterSpec	Provides Parameters for setting up a Databricks Cluster
JobCluster	Sets up a Jobs Cluster as defined by the provided `DatabricksCluster`
Task	Defines the setup of the Task at the Databricks Task level including Task specific Clusters, Libraries, Schedules, Notifications and Timeouts
CreateJob	Defines the setup at the Job level including Clusters, Libraries, Schedules, Notifications, Access Controls, Timeouts and Tags
NotebookTask	Provides the Notebook information to the `Task`
DatabricksSDKDeploy	Leverages the Databricks SDK to deploy the job to Databricks Workflows

Note

All classes for deployment are available from the Databricks SDK and can be accessed using from rtdip_sdk.pipelines.deploy import and choosing the classes you need for your Databricks deployment

Example

A simple example of deploying an RTDIP Pipeline Job to an Azure Databricks Job is below.

databricks_host_name = "{databricks-host-url}" #Replace with your databricks workspace url

# Setup a Cluster for the Databricks Job
cluster_list = []
cluster_list.append(JobCluster(
    job_cluster_key="test_cluster",
    new_cluster=ClusterSpec(
        node_type_id="Standard_E4ds_v5",
        autoscale=AutoScale(min_workers=1, max_workers=3),
        spark_version="13.2.x-scala2.12",
        data_security_mode=DataSecurityMode.SINGLE_USER,
        runtime_engine=RuntimeEngine.PHOTON
    )
))

# Define a Notebook Task for the Databricks Job
task_list = []
task_list.append(Task(
    task_key="test_task",
    job_cluster_key="test_cluster",
    notebook_task=NotebookTask(
        notebook_path="/directory/to/pipeline.py"
    )
))

# Create a Databricks Job for the Task
job = CreateJob(
    name="test_job_rtdip",
    job_clusters=cluster_list,
    tasks=task_list
)

# Deploy to Databricks
databricks_job = DatabricksSDKDeploy(databricks_job=job, host=databricks_host_name, token=access_token)

deploy_result = databricks_job.deploy()

Launch

Once a job is deployed to Databricks, it can be executed immediately using the following code.

# Run/Launch the Job in Databricks
launch_result = databricks_job.launch()

Stop

A job that is running and is deployed to Databricks, can be cancelled using the following code.

```python

Run/Launch the Job in Databricks

stop_result = databricks_job.stop()