Databricks miner package

The Databricks miner package extract lineage and usage from databricks to Atlan for discovery.

Will create a new connection

This should only be used to create the workflow the first time. Each time you run this method it will create a new connection and new assets within that connection — which could lead to duplicate assets if you run the workflow this way multiple times with the same settings.

Instead, when you want to re-crawl assets, re-run the existing workflow (see Re-run existing workflow below).

4.0.1

To extract lineage and usage from databricks to Atlan for discovery.

Java Python Kotlin Raw REST API

Coming soon

from pyatlan.client.atlan import AtlanClient
from pyatlan.cache.role_cache import RoleCache
from pyatlan.model.packages import DatabricksMiner

client = AtlanClient()

crawler = (
    DatabricksMiner( # (1)
        connection_qualified_name="default/databricks/1234567890"
        # (2)
    )
    .rest_api() # (3)
    .popularity_configuration( # (4)
        start_date="1234567890",
        extraction_method=DatabricksMiner.ExtractionMethod.SYSTEM_TABLE,
        window_days=30,
        excluded_users=["test-user-1", "test-user-2"],
        warehouse_id="test-warehouse-id",
    )
    .to_workflow() # (5)
)
response = client.workflow.run(crawler) # (6)

1. Base configuration for a new Databricks miner.
2. You must provide the exact qualified_name of the Databricks connection in Atlan for which you want to mine query history.

3. You can sets up the Databricks miner to use the REST API method for fetching lineage.

   You can also utilize any of the following methods for fetching lineage:

   • offline()

     • bucket_name: name of the S3 bucket to extract data from.
     • bucket_prefix: prefix within the S3 bucket to narrow the extraction scope.

   • system_table()

     • warehouse_id: unique identifier of the SQL warehouse to be used for system table extraction.

4. Optionally, you can define popularity_configuration():

   • epoch timestamp from which queries will be fetched for calculating popularity. This does not affect lineage generation.
   • method used to fetch popularity data. Defaults to ExtractionMethod.REST_API.
   • (Optional) number of days to consider for calculating popularity metrics.
   • (Optional) list of usernames to exclude from usage metrics calculations.
   • (Optional) unique identifier of the SQL warehouse to use for popularity calculations. Required if extraction_method is ExtractionMethod.SYSTEM_TABLE.

5. Now, you can convert the package into a Workflow object.

6. Run the workflow by invoking the run() method on the workflow client, passing the created object.

   Workflows run asynchronously

   Remember that workflows run asynchronously. See the packages and workflows introduction for details on how you can check the status and wait until the workflow has been completed.

Coming soon

Create the workflow via UI only

We recommend creating the workflow only via the UI. To rerun an existing workflow, see the steps below.

Re-run existing workflow

1.1.0

To re-run an existing workflow for databricks assets:

Java Python Kotlin Raw REST API

Coming soon

from pyatlan.client.atlan import AtlanClient
from pyatlan.model.enums import WorkflowPackage

client = AtlanClient()

existing = client.workflow.find_by_type(  # (1)
  prefix=WorkflowPackage.DATABRICKS_LINEAGE, max_results=5
)

# Determine which Databricks workflow (n)
# from the list of results you want to re-run.
response = client.workflow.rerun(existing[n]) # (2)

1. You can find workflows by their type using the workflow client find_by_type() method and providing the prefix for one of the packages. In this example, we do so for the DatabricksMiner. (You can also specify the maximum number of resulting workflows you want to retrieve as results.)

2. Once you've found the workflow you want to re-run, you can simply call the workflow client rerun() method.

   • Optionally, you can use rerun(idempotent=True) to avoid re-running a workflow that is already in running or in a pending state. This will return details of the already running workflow if found, and by default, it is set to False.

   Workflows run asynchronously

   Remember that workflows run asynchronously. See the packages and workflows introduction for details on how you can check the status and wait until the workflow has been completed.

Coming soon

Requires multiple steps through the raw REST API

1. Find the existing workflow.
2. Send through the resulting re-run request.

{
  "from": 0,
  "size": 5,
  "query": {
    "bool": {
      "filter": [
        {
          "nested": {
            "path": "metadata",
            "query": {
              "prefix": {
                "metadata.name.keyword": {
                  "value": "atlan-databricks-lineage" // (1)
                }
              }
            }
          }
        }
      ]
    }
  },
  "sort": [
    {
      "metadata.creationTimestamp": {
        "nested": {
          "path": "metadata"
        },
        "order": "desc"
      }
    }
  ],
  "track_total_hits": true
}

1. Searching by the atlan-databricks-lineage prefix will ensure you only find existing Databricks assets workflows.

   Name of the workflow

   The name of the workflow will be nested within the _source.metadata.name property of the response object. (Remember since this is a search, there could be multiple results, so you may want to use the other details in each result to determine which workflow you really want.)

{
  "namespace": "default",
  "resourceKind": "WorkflowTemplate",
  "resourceName": "atlan-databricks-lineage-1684500411" // (1)
}

1. Send the name of the workflow as the resourceName to rerun it.

---