Scans and ingestion in Data Map

This article provides an overview of the scanning and ingestion features in Microsoft Purview Data Map. These features connect your Microsoft Purview account to your sources to populate Data Map and Unified Catalog so you can begin exploring and managing your data through Microsoft Purview.

• Scanning captures metadata from data sources and brings it to Microsoft Purview.
• Ingestion processes metadata and stores it in Unified Catalog from both:
  • Data source scans - scanned metadata is added to Data Map.
  • Lineage connections - transformation resources add metadata about their sources, outputs, and activities to Data Map.

Scanning

After you register data sources in your Microsoft Purview account, scan the data sources. The scanning process connects to the data source and captures technical metadata like names, file size, columns, and more. It also extracts schema for structured data sources, applies classifications on schemas. You can trigger the scanning process to run immediately or schedule it to run periodically keep your Microsoft Purview account up to date.

For each scan, you can customize the process so that you only scan the information you need, rather than the whole source.

Choose an authentication method for your scans

Microsoft Purview is secure by default. It doesn't store passwords or secrets directly, so you need to choose an authentication method for your sources. You can authenticate your Microsoft Purview account in several ways, but not all methods are supported for each data source.

• Managed Identity
• Service Principal
• SQL Authentication
• Windows Authentication
• Role ARN
• Delegated Authentication
• Consumer Key
• Account Key or Basic Authentication

Whenever possible, use a Managed Identity because it eliminates the need for storing and managing credentials for individual data sources. This method can greatly reduce the time you and your team spend setting up and troubleshooting authentication for scans. When you enable a managed identity for your Microsoft Purview account, an identity is created in Microsoft Entra ID (Entra ID) and is tied to the lifecycle of your account.

Scope your scan

When scanning a source, you can scan the entire data source or choose only specific entities (folders or tables) to scan. The available options depend on the source you're scanning. You can define these options for both one-time and scheduled scans.

For example, when creating and running a scan for an Azure SQL Database, you can choose which tables to scan, or select the entire database.

For each entity (folder or table), three selection states exist: fully selected, partially selected, and not selected. In the following example, if you select Department 1 on the folder hierarchy, Department 1 is considered as fully selected. The parent entities for Department 1, which are Company and example, are considered as partially selected, since the other entities under the same parent (for example, Department 2) aren't selected. Different icons are used on the UI for entities with different selection states.

After you run the scan, it's likely that new assets will be added in the source system. By default, future assets under a certain parent are automatically selected if the parent is fully or partially selected when you run the scan again. In the preceding example, after you select Department 1 and run the scan, any new assets under folder Department 1 or under Company and example are included when you run the scan again.

A toggle button, as seen in the image below, allows users to control the automatic inclusion for new assets under partially selected parent. By default, the toggle is turned off and the automatic inclusion behavior for a partially selected parent is disabled. With the toggle turned off, any new assets under partially selected parents like Company and example aren't included when you run the scan again; only new assets under Department 1 are included in future scan.

If the toggle is turned on, the new assets under a parent are automatically selected if the parent is fully or partially selected when you run the scan again. The inclusion behavior is the same as before the toggle is introduced.

Known limitations

When the toggle is turned off:

• The file entities under a partially selected parent aren't scanned.
• If all existing entities under a parent are explicitly selected, the parent is considered fully selected, and any new assets under the parent are included when you run the scan again.

Scan rule set

A scan rule set determines the kinds of information a scan looks for when it's running against one of your sources. Available rules depend on the kind of source you're scanning, but include things like the file types you should scan, and the kinds of classifications you need.

Many data source types already have system scan rule sets, but you can also create your own scan rule sets to tailor your scans to your organization.

Schedule your scan

Microsoft Purview gives you a choice of scanning daily, weekly, or monthly at a specific time you choose. Learn more about the supported schedule options. Daily or weekly scans might be appropriate for data sources with structures that are actively under development or frequently change. Monthly scanning is more appropriate for data sources that change infrequently. Work with the administrator of the source you want to scan to identify a time when compute demands on the source are low.

How scans detect deleted assets

A Microsoft Purview catalog is only aware of the state of a data store when it runs a scan. For the catalog to know if a file, table, or container is deleted, it compares the last scan output against the current scan output. For example, suppose that the last time you scanned an Azure Data Lake Storage Gen2 account, it included a folder named folder1. When the same account is scanned again, folder1 is missing. Therefore, the catalog assumes the folder is deleted.

Detecting deleted files

The logic for detecting missing files works for multiple scans by the same user and by different users. For example, suppose a user runs a one-time scan on a Data Lake Storage Gen2 data store on folders A, B, and C. Later, a different user in the same account runs a different one-time scan on folders C, D, and E of the same data store. Because folder C was scanned twice, the catalog checks it for possible deletions. Folders A, B, D, and E, however, were scanned only once, and the catalog doesn't check them for deleted assets.

To keep deleted files out of your catalog, it's important to run regular scans. The scan interval is important, because the catalog can't detect deleted assets until another scan is run. So, if you run scans once a month on a particular store, the catalog can't detect any deleted data assets in that store until you run the next scan a month later.

When you enumerate large data stores like Data Lake Storage Gen2, there are multiple ways (including enumeration errors and dropped events) to miss information. A particular scan might miss that a file was created or deleted. So, unless the catalog is certain a file is deleted, it doesn't delete it from the catalog. This strategy means there can be errors when a file that doesn't exist in the scanned data store still exists in the catalog. In some cases, a data store might need to be scanned two or three times before it catches certain deleted assets.

Ingestion

Ingestion is the process that populates Data Map with metadata gathered through its various processes.

Next steps

For more information, or for specific instructions for scanning sources, follow the links below.

• Understanding resource sets
• How to govern an Azure SQL Database