Copy and transform data in Azure Data Lake Storage Gen2 using Azure Data Factory or Azure Synapse Analytics

APPLIES TO: Azure Data Factory Azure Synapse Analytics

Tip

Try out Data Factory in Microsoft Fabric, an all-in-one analytics solution for enterprises. Microsoft Fabric covers everything from data movement to data science, real-time analytics, business intelligence, and reporting. Learn how to start a new trial for free!

Azure Data Lake Storage Gen2 (ADLS Gen2) is a set of capabilities dedicated to big data analytics built into Azure Blob storage. You can use it to interface with your data by using both file system and object storage paradigms.

This article outlines how to use Copy Activity to copy data from and to Azure Data Lake Storage Gen2, and use Data Flow to transform data in Azure Data Lake Storage Gen2. To learn more, read the introductory article for Azure Data Factory or Azure Synapse Analytics.

Tip

For data lake or data warehouse migration scenario, learn more in Migrate data from your data lake or data warehouse to Azure.

Supported capabilities

This Azure Data Lake Storage Gen2 connector is supported for the following capabilities:

Supported capabilities IR Managed private endpoint
Copy activity (source/sink) ① ②
Mapping data flow (source/sink)
Lookup activity ① ②
GetMetadata activity ① ②
Delete activity ① ②

① Azure integration runtime ② Self-hosted integration runtime

For Copy activity, with this connector you can:

Get started

Tip

For a walk-through of how to use the Data Lake Storage Gen2 connector, see Load data into Azure Data Lake Storage Gen2.

To perform the Copy activity with a pipeline, you can use one of the following tools or SDKs:

Create an Azure Data Lake Storage Gen2 linked service using UI

Use the following steps to create an Azure Data Lake Storage Gen2 linked service in the Azure portal UI.

  1. Browse to the Manage tab in your Azure Data Factory or Synapse workspace and select Linked Services, then click New:

  2. Search for Azure Data Lake Storage Gen2 and select the Azure Data Lake Storage Gen2 connector.

    Select Azure Data Lake Storage Gen2 connector.

  3. Configure the service details, test the connection, and create the new linked service.

    Screenshot of configuration for Azure Data Lake Storage Gen2 linked service.

Connector configuration details

The following sections provide information about properties that are used to define Data Factory and Synapse pipeline entities specific to Data Lake Storage Gen2.

Linked service properties

The Azure Data Lake Storage Gen2 connector supports the following authentication types. See the corresponding sections for details:

Note

  • If want to use the public Azure integration runtime to connect to the Data Lake Storage Gen2 by leveraging the Allow trusted Microsoft services to access this storage account option enabled on Azure Storage firewall, you must use managed identity authentication. For more information about the Azure Storage firewalls settings, see Configure Azure Storage firewalls and virtual networks.
  • When you use PolyBase or COPY statement to load data into Azure Synapse Analytics, if your source or staging Data Lake Storage Gen2 is configured with an Azure Virtual Network endpoint, you must use managed identity authentication as required by Azure Synapse. See the managed identity authentication section with more configuration prerequisites.

Account key authentication

To use storage account key authentication, the following properties are supported:

Property Description Required
type The type property must be set to AzureBlobFS. Yes
url Endpoint for Data Lake Storage Gen2 with the pattern of https://<accountname>.dfs.core.chinacloudapi.cn. Yes
accountKey Account key for Data Lake Storage Gen2. Mark this field as a SecureString to store it securely, or reference a secret stored in Azure Key Vault. Yes
connectVia The integration runtime to be used to connect to the data store. You can use the Azure integration runtime or a self-hosted integration runtime if your data store is in a private network. If this property isn't specified, the default Azure integration runtime is used. No

Note

Secondary ADLS file system endpoint is not supported when using account key authentication. You can use other authentication types.

Example:

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "url": "https://<accountname>.dfs.core.chinacloudapi.cn", 
            "accountkey": { 
                "type": "SecureString", 
                "value": "<accountkey>" 
            }
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

Shared access signature authentication

A shared access signature provides delegated access to resources in your storage account. You can use a shared access signature to grant a client limited permissions to objects in your storage account for a specified time.

You don't have to share your account access keys. The shared access signature is a URI that encompasses in its query parameters all the information necessary for authenticated access to a storage resource. To access storage resources with the shared access signature, the client only needs to pass in the shared access signature to the appropriate constructor or method.

For more information about shared access signatures, see Shared access signatures: Understand the shared access signature model.

Note

  • The service now supports both service shared access signatures and account shared access signatures. For more information about shared access signatures, see Grant limited access to Azure Storage resources using shared access signatures.
  • In later dataset configurations, the folder path is the absolute path starting from the container level. You need to configure one aligned with the path in your SAS URI.

The following properties are supported for using shared access signature authentication:

Property Description Required
type The type property must be set to AzureBlobFS (suggested) Yes
sasUri Specify the shared access signature URI to the Storage resources such as blob or container.
Mark this field as SecureString to store it securely. You can also put the SAS token in Azure Key Vault to use auto-rotation and remove the token portion. For more information, see the following samples and Store credentials in Azure Key Vault.
Yes
connectVia The integration runtime to be used to connect to the data store. You can use the Azure integration runtime or the self-hosted integration runtime (if your data store is in a private network). If this property isn't specified, the service uses the default Azure integration runtime. No

Note

If you're using the AzureStorage type linked service, it's still supported as is. But we suggest that you use the new AzureDataLakeStorageGen2 linked service type going forward.

Example:

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "sasUri": {
                "type": "SecureString",
                "value": "<SAS URI of the Azure Storage resource e.g. https://<accountname>.blob.core.chinacloudapi.cn/?sv=<storage version>&st=<start time>&se=<expire time>&sr=<resource>&sp=<permissions>&sip=<ip range>&spr=<protocol>&sig=<signature>>"
            }
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

Example: store the account key in Azure Key Vault

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "sasUri": {
                "type": "SecureString",
                "value": "<SAS URI of the Azure Storage resource without token e.g. https://<accountname>.blob.core.chinacloudapi.cn/>"
            },
            "sasToken": {
                "type": "AzureKeyVaultSecret",
                "store": {
                    "referenceName": "<Azure Key Vault linked service name>", 
                    "type": "LinkedServiceReference"
                },
                "secretName": "<secretName with value of SAS token e.g. ?sv=<storage version>&st=<start time>&se=<expire time>&sr=<resource>&sp=<permissions>&sip=<ip range>&spr=<protocol>&sig=<signature>>"
            }
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

When you create a shared access signature URI, consider the following points:

  • Set appropriate read/write permissions on objects based on how the linked service (read, write, read/write) is used.
  • Set Expiry time appropriately. Make sure that the access to Storage objects doesn't expire within the active period of the pipeline.
  • The URI should be created at the right container or blob based on the need. A shared access signature URI to a blob allows the data factory or Synapse pipeline to access that particular blob. A shared access signature URI to a Blob storage container allows the data factory or Synapse pipeline to iterate through blobs in that container. To provide access to more or fewer objects later, or to update the shared access signature URI, remember to update the linked service with the new URI.

Service principal authentication

To use service principal authentication, follow these steps.

  1. Register an application with the Microsoft identity platform. To learn how, see Quickstart: Register an application with the Microsoft identity platform. Make note of these values, which you use to define the linked service:

    • Application ID
    • Application key
    • Tenant ID
  2. Grant the service principal proper permission. See examples on how permission works in Data Lake Storage Gen2 from Access control lists on files and directories

    • As source: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Read permission for the files to copy. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Reader role.
    • As sink: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Write permission for the sink folder. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Contributor role.

Note

If you use UI to author and the service principal is not set with "Storage Blob Data Reader/Contributor" role in IAM, when doing test connection or browsing/navigating folders, choose "Test connection to file path" or "Browse from specified path", and specify a path with Read + Execute permission to continue.

These properties are supported for the linked service:

Property Description Required
type The type property must be set to AzureBlobFS. Yes
url Endpoint for Data Lake Storage Gen2 with the pattern of https://<accountname>.dfs.core.chinacloudapi.cn. Yes
servicePrincipalId Specify the application's client ID. Yes
servicePrincipalCredentialType The credential type to use for service principal authentication. Allowed values are ServicePrincipalKey and ServicePrincipalCert. Yes
servicePrincipalCredential The service principal credential.
When you use ServicePrincipalKey as the credential type, specify the application's key. Mark this field as SecureString to store it securely, or reference a secret stored in Azure Key Vault.
When you use ServicePrincipalCert as the credential, reference a certificate in Azure Key Vault, and ensure the certificate content type is PKCS #12.
Yes
servicePrincipalKey Specify the application's key. Mark this field as SecureString to store it securely, or reference a secret stored in Azure Key Vault.
This property is still supported as-is for servicePrincipalId + servicePrincipalKey. As ADF adds new service principal certificate authentication, the new model for service principal authentication is servicePrincipalId + servicePrincipalCredentialType + servicePrincipalCredential.
No
tenant Specify the tenant information (domain name or tenant ID) under which your application resides. Retrieve it by hovering the mouse in the upper-right corner of the Azure portal. Yes
azureCloudType For service principal authentication, specify the type of Azure cloud environment to which your Microsoft Entra application is registered.
Allowed value is AzureChina. By default, the data factory or or Synapse pipeline's cloud environment is used.
No
connectVia The integration runtime to be used to connect to the data store. You can use the Azure integration runtime or a self-hosted integration runtime if your data store is in a private network. If not specified, the default Azure integration runtime is used. No

Example: using service principal key authentication

You can also store service principal key in Azure Key Vault.

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "url": "https://<accountname>.dfs.core.chinacloudapi.cn", 
            "servicePrincipalId": "<service principal id>",
            "servicePrincipalCredentialType": "ServicePrincipalKey",
            "servicePrincipalCredential": {
                "type": "SecureString",
                "value": "<service principal key>"
            },
            "tenant": "<tenant info, e.g. microsoft.partner.onmschina.cn>" 
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

Example: using service principal certificate authentication

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "url": "https://<accountname>.dfs.core.chinacloudapi.cn", 
            "servicePrincipalId": "<service principal id>",
            "servicePrincipalCredentialType": "ServicePrincipalCert",
            "servicePrincipalCredential": { 
                "type": "AzureKeyVaultSecret", 
                "store": { 
                    "referenceName": "<AKV reference>", 
                    "type": "LinkedServiceReference" 
                }, 
                "secretName": "<certificate name in AKV>" 
            },
            "tenant": "<tenant info, e.g. microsoft.partner.onmschina.cn>" 
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

System-assigned managed identity authentication

A data factory or Synapse workspace can be associated with a system-assigned managed identity. You can directly use this system-assigned managed identity for Data Lake Storage Gen2 authentication, similar to using your own service principal. It allows this designated factory or workspace to access and copy data to or from your Data Lake Storage Gen2.

To use system-assigned managed identity authentication, follow these steps.

  1. Retrieve the system-assigned managed identity information by copying the value of the managed identity object ID generated along with your data factory or Synapse workspace.

  2. Grant the system-assigned managed identity proper permission. See examples on how permission works in Data Lake Storage Gen2 from Access control lists on files and directories.

    • As source: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Read permission for the files to copy. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Reader role.
    • As sink: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Write permission for the sink folder. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Contributor role.

These properties are supported for the linked service:

Property Description Required
type The type property must be set to AzureBlobFS. Yes
url Endpoint for Data Lake Storage Gen2 with the pattern of https://<accountname>.dfs.core.chinacloudapi.cn. Yes
connectVia The integration runtime to be used to connect to the data store. You can use the Azure integration runtime or a self-hosted integration runtime if your data store is in a private network. If not specified, the default Azure integration runtime is used. No

Example:

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "url": "https://<accountname>.dfs.core.chinacloudapi.cn", 
        },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

User-assigned managed identity authentication

A data factory can be assigned with one or multiple user-assigned managed identities. You can use this user-assigned managed identity for Blob storage authentication, which allows to access and copy data from or to Data Lake Storage Gen2. To learn more about managed identities for Azure resources, see Managed identities for Azure resources

To use user-assigned managed identity authentication, follow these steps:

  1. Create one or multiple user-assigned managed identities and grant access to Azure Data Lake Storage Gen2. See examples on how permission works in Data Lake Storage Gen2 from Access control lists on files and directories.

    • As source: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Read permission for the files to copy. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Reader role.
    • As sink: In Storage Explorer, grant at least Execute permission for ALL upstream folders and the file system, along with Write permission for the sink folder. Alternatively, in Access control (IAM), grant at least the Storage Blob Data Contributor role.
  2. Assign one or multiple user-assigned managed identities to your data factory and create credentials for each user-assigned managed identity.

These properties are supported for the linked service:

Property Description Required
type The type property must be set to AzureBlobFS. Yes
url Endpoint for Data Lake Storage Gen2 with the pattern of https://<accountname>.dfs.core.chinacloudapi.cn. Yes
credentials Specify the user-assigned managed identity as the credential object. Yes
connectVia The integration runtime to be used to connect to the data store. You can use the Azure integration runtime or a self-hosted integration runtime if your data store is in a private network. If not specified, the default Azure integration runtime is used. No

Example:

{
    "name": "AzureDataLakeStorageGen2LinkedService",
    "properties": {
        "type": "AzureBlobFS",
        "typeProperties": {
            "url": "https://<accountname>.dfs.core.chinacloudapi.cn", 
            "credential": {
                "referenceName": "credential1",
                "type": "CredentialReference"
                }
            },
        "connectVia": {
            "referenceName": "<name of Integration Runtime>",
            "type": "IntegrationRuntimeReference"
        }
    }
}

Note

If you use Data Factory UI to author and the managed identity is not set with "Storage Blob Data Reader/Contributor" role in IAM, when doing test connection or browsing/navigating folders, choose "Test connection to file path" or "Browse from specified path", and specify a path with Read + Execute permission to continue.

Important

If you use PolyBase or COPY statement to load data from Data Lake Storage Gen2 into Azure Synapse Analytics, when you use managed identity authentication for Data Lake Storage Gen2, make sure you also follow steps 1 to 3 in this guidance. Those steps will register your server with Microsoft Entra ID and assign the Storage Blob Data Contributor role to your server. Data Factory handles the rest. If you configure Blob storage with an Azure Virtual Network endpoint, you also need to have Allow trusted Microsoft services to access this storage account turned on under Azure Storage account Firewalls and Virtual networks settings menu as required by Azure Synapse.

Dataset properties

For a full list of sections and properties available for defining datasets, see Datasets.

Azure Data Factory supports the following file formats. Refer to each article for format-based settings.

The following properties are supported for Data Lake Storage Gen2 under location settings in the format-based dataset:

Property Description Required
type The type property under location in the dataset must be set to AzureBlobFSLocation. Yes
fileSystem The Data Lake Storage Gen2 file system name. No
folderPath The path to a folder under the given file system. If you want to use a wildcard to filter folders, skip this setting and specify it in activity source settings. No
fileName The file name under the given fileSystem + folderPath. If you want to use a wildcard to filter files, skip this setting and specify it in activity source settings. No

Example:

{
    "name": "DelimitedTextDataset",
    "properties": {
        "type": "DelimitedText",
        "linkedServiceName": {
            "referenceName": "<Data Lake Storage Gen2 linked service name>",
            "type": "LinkedServiceReference"
        },
        "schema": [ < physical schema, optional, auto retrieved during authoring > ],
        "typeProperties": {
            "location": {
                "type": "AzureBlobFSLocation",
                "fileSystem": "filesystemname",
                "folderPath": "folder/subfolder"
            },
            "columnDelimiter": ",",
            "quoteChar": "\"",
            "firstRowAsHeader": true,
            "compressionCodec": "gzip"
        }
    }
}

Copy activity properties

For a full list of sections and properties available for defining activities, see Copy activity configurations and Pipelines and activities. This section provides a list of properties supported by the Data Lake Storage Gen2 source and sink.

Azure Data Lake Storage Gen2 as a source type

Azure Data Factory supports the following file formats. Refer to each article for format-based settings.

You have several options to copy data from ADLS Gen2:

  • Copy from the given path specified in the dataset.
  • Wildcard filter against folder path or file name, see wildcardFolderPath and wildcardFileName.
  • Copy the files defined in a given text file as file set, see fileListPath.

The following properties are supported for Data Lake Storage Gen2 under storeSettings settings in format-based copy source:

Property Description Required
type The type property under storeSettings must be set to AzureBlobFSReadSettings. Yes
Locate the files to copy:
OPTION 1: static path
Copy from the given file system or folder/file path specified in the dataset. If you want to copy all files from a file system/folder, additionally specify wildcardFileName as *.
OPTION 2: wildcard
- wildcardFolderPath
The folder path with wildcard characters under the given file system configured in dataset to filter source folders.
Allowed wildcards are: * (matches zero or more characters) and ? (matches zero or single character); use ^ to escape if your actual folder name has wildcard or this escape char inside.
See more examples in Folder and file filter examples.
No
OPTION 2: wildcard
- wildcardFileName
The file name with wildcard characters under the given file system + folderPath/wildcardFolderPath to filter source files.
Allowed wildcards are: * (matches zero or more characters) and ? (matches zero or single character); use ^ to escape if your actual file name has wildcard or this escape char inside. See more examples in Folder and file filter examples.
Yes
OPTION 3: a list of files
- fileListPath
Indicates to copy a given file set. Point to a text file that includes a list of files you want to copy, one file per line, which is the relative path to the path configured in the dataset.
When using this option, do not specify file name in dataset. See more examples in File list examples.
No
Additional settings:
recursive Indicates whether the data is read recursively from the subfolders or only from the specified folder. Note that when recursive is set to true and the sink is a file-based store, an empty folder or subfolder isn't copied or created at the sink.
Allowed values are true (default) and false.
This property doesn't apply when you configure fileListPath.
No
deleteFilesAfterCompletion Indicates whether the binary files will be deleted from source store after successfully moving to the destination store. The file deletion is per file, so when copy activity fails, you will see some files have already been copied to the destination and deleted from source, while others are still remaining on source store.
This property is only valid in binary files copy scenario. The default value: false.
No
modifiedDatetimeStart Files filter based on the attribute: Last Modified.
The files will be selected if their last modified time is greater than or equal to modifiedDatetimeStart and less than modifiedDatetimeEnd. The time is applied to UTC time zone in the format of "2018-12-01T05:00:00Z".
The properties can be NULL, which means no file attribute filter will be applied to the dataset. When modifiedDatetimeStart has datetime value but modifiedDatetimeEnd is NULL, it means the files whose last modified attribute is greater than or equal with the datetime value will be selected. When modifiedDatetimeEnd has datetime value but modifiedDatetimeStart is NULL, it means the files whose last modified attribute is less than the datetime value will be selected.
This property doesn't apply when you configure fileListPath.
No
modifiedDatetimeEnd Same as above. No
enablePartitionDiscovery For files that are partitioned, specify whether to parse the partitions from the file path and add them as additional source columns.
Allowed values are false (default) and true.
No
partitionRootPath When partition discovery is enabled, specify the absolute root path in order to read partitioned folders as data columns.

If it is not specified, by default,
- When you use file path in dataset or list of files on source, partition root path is the path configured in dataset.
- When you use wildcard folder filter, partition root path is the sub-path before the first wildcard.

For example, assuming you configure the path in dataset as "root/folder/year=2020/month=08/day=27":
- If you specify partition root path as "root/folder/year=2020", copy activity will generate two more columns month and day with value "08" and "27" respectively, in addition to the columns inside the files.
- If partition root path is not specified, no extra column will be generated.
No
maxConcurrentConnections The upper limit of concurrent connections established to the data store during the activity run. Specify a value only when you want to limit concurrent connections. No

Example:

"activities":[
    {
        "name": "CopyFromADLSGen2",
        "type": "Copy",
        "inputs": [
            {
                "referenceName": "<Delimited text input dataset name>",
                "type": "DatasetReference"
            }
        ],
        "outputs": [
            {
                "referenceName": "<output dataset name>",
                "type": "DatasetReference"
            }
        ],
        "typeProperties": {
            "source": {
                "type": "DelimitedTextSource",
                "formatSettings":{
                    "type": "DelimitedTextReadSettings",
                    "skipLineCount": 10
                },
                "storeSettings":{
                    "type": "AzureBlobFSReadSettings",
                    "recursive": true,
                    "wildcardFolderPath": "myfolder*A",
                    "wildcardFileName": "*.csv"
                }
            },
            "sink": {
                "type": "<sink type>"
            }
        }
    }
]

Azure Data Lake Storage Gen2 as a sink type

Azure Data Factory supports the following file formats. Refer to each article for format-based settings.

The following properties are supported for Data Lake Storage Gen2 under storeSettings settings in format-based copy sink:

Property Description Required
type The type property under storeSettings must be set to AzureBlobFSWriteSettings. Yes
copyBehavior Defines the copy behavior when the source is files from a file-based data store.

Allowed values are:
- PreserveHierarchy (default): Preserves the file hierarchy in the target folder. The relative path of the source file to the source folder is identical to the relative path of the target file to the target folder.
- FlattenHierarchy: All files from the source folder are in the first level of the target folder. The target files have autogenerated names.
- MergeFiles: Merges all files from the source folder to one file. If the file name is specified, the merged file name is the specified name. Otherwise, it's an autogenerated file name.
No
blockSizeInMB Specify the block size in MB used to write data to ADLS Gen2. Learn more about Block Blobs.
Allowed value is between 4 MB and 100 MB.
By default, ADF automatically determines the block size based on your source store type and data. For non-binary copy into ADLS Gen2, the default block size is 100 MB so as to fit in at most approximately 4.75-TB data. It may be not optimal when your data is not large, especially when you use Self-hosted Integration Runtime with poor network resulting in operation timeout or performance issue. You can explicitly specify a block size, while ensure blockSizeInMB*50000 is big enough to store the data, otherwise copy activity run will fail.
No
maxConcurrentConnections The upper limit of concurrent connections established to the data store during the activity run. Specify a value only when you want to limit concurrent connections. No
metadata Set custom metadata when copy to sink. Each object under the metadata array represents an extra column. The name defines the metadata key name, and the value indicates the data value of that key. If preserve attributes feature is used, the specified metadata will union/overwrite with the source file metadata.

Allowed data values are:
- $$LASTMODIFIED: a reserved variable indicates to store the source files' last modified time. Apply to file-based source with binary format only.
- Expression
- Static value
No

Example:

"activities":[
    {
        "name": "CopyToADLSGen2",
        "type": "Copy",
        "inputs": [
            {
                "referenceName": "<input dataset name>",
                "type": "DatasetReference"
            }
        ],
        "outputs": [
            {
                "referenceName": "<Parquet output dataset name>",
                "type": "DatasetReference"
            }
        ],
        "typeProperties": {
            "source": {
                "type": "<source type>"
            },
            "sink": {
                "type": "ParquetSink",
                "storeSettings":{
                    "type": "AzureBlobFSWriteSettings",
                    "copyBehavior": "PreserveHierarchy",
                    "metadata": [
                        {
                            "name": "testKey1",
                            "value": "value1"
                        },
                        {
                            "name": "testKey2",
                            "value": "value2"
                        },
                        {
                            "name": "lastModifiedKey",
                            "value": "$$LASTMODIFIED"
                        }
                    ]
                }
            }
        }
    }
]

Folder and file filter examples

This section describes the resulting behavior of the folder path and file name with wildcard filters.

folderPath fileName recursive Source folder structure and filter result (files in bold are retrieved)
Folder* (Empty, use default) false FolderA
    File1.csv
    File2.json
    Subfolder1
        File3.csv
        File4.json
        File5.csv
AnotherFolderB
    File6.csv
Folder* (Empty, use default) true FolderA
    File1.csv
    File2.json
    Subfolder1
        File3.csv
        File4.json
        File5.csv
AnotherFolderB
    File6.csv
Folder* *.csv false FolderA
    File1.csv
    File2.json
    Subfolder1
        File3.csv
        File4.json
        File5.csv
AnotherFolderB
    File6.csv
Folder* *.csv true FolderA
    File1.csv
    File2.json
    Subfolder1
        File3.csv
        File4.json
        File5.csv
AnotherFolderB
    File6.csv

File list examples

This section describes the resulting behavior of using file list path in copy activity source.

Assuming you have the following source folder structure and want to copy the files in bold:

Sample source structure Content in FileListToCopy.txt ADF configuration
filesystem
    FolderA
        File1.csv
        File2.json
        Subfolder1
            File3.csv
            File4.json
            File5.csv
    Metadata
        FileListToCopy.txt
File1.csv
Subfolder1/File3.csv
Subfolder1/File5.csv
In dataset:
- File system: filesystem
- Folder path: FolderA

In copy activity source:
- File list path: filesystem/Metadata/FileListToCopy.txt

The file list path points to a text file in the same data store that includes a list of files you want to copy, one file per line with the relative path to the path configured in the dataset.

Some recursive and copyBehavior examples

This section describes the resulting behavior of the copy operation for different combinations of recursive and copyBehavior values.

recursive copyBehavior Source folder structure Resulting target
true preserveHierarchy Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the same structure as the source:

Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
true flattenHierarchy Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the following structure:

Folder1
    autogenerated name for File1
    autogenerated name for File2
    autogenerated name for File3
    autogenerated name for File4
    autogenerated name for File5
true mergeFiles Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the following structure:

Folder1
    File1 + File2 + File3 + File4 + File5 contents are merged into one file with an autogenerated file name.
false preserveHierarchy Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the following structure:

Folder1
    File1
    File2

Subfolder1 with File3, File4, and File5 isn't picked up.
false flattenHierarchy Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the following structure:

Folder1
    autogenerated name for File1
    autogenerated name for File2

Subfolder1 with File3, File4, and File5 isn't picked up.
false mergeFiles Folder1
    File1
    File2
    Subfolder1
        File3
        File4
        File5
The target Folder1 is created with the following structure:

Folder1
    File1 + File2 contents are merged into one file with an autogenerated file name. autogenerated name for File1

Subfolder1 with File3, File4, and File5 isn't picked up.

Preserve metadata during copy

When you copy files from Amazon S3/Azure Blob/Azure Data Lake Storage Gen2 to Azure Data Lake Storage Gen2/Azure Blob, you can choose to preserve the file metadata along with data. Learn more from Preserve metadata.

Mapping data flow properties

When you're transforming data in mapping data flows, you can read and write files from Azure Data Lake Storage Gen2 in the following formats:

Format specific settings are located in the documentation for that format. For more information, see Source transformation in mapping data flow and Sink transformation in mapping data flow.

Source transformation

In the source transformation, you can read from a container, folder, or individual file in Azure Data Lake Storage Gen2. The Source options tab lets you manage how the files get read.

Screenshot of source options tab in mapping data flow source transformation.

Wildcard path: Using a wildcard pattern will instruct ADF to loop through each matching folder and file in a single Source transformation. This is an effective way to process multiple files within a single flow. Add multiple wildcard matching patterns with the + sign that appears when hovering over your existing wildcard pattern.

From your source container, choose a series of files that match a pattern. Only container can be specified in the dataset. Your wildcard path must therefore also include your folder path from the root folder.

Wildcard examples:

  • * Represents any set of characters

  • ** Represents recursive directory nesting

  • ? Replaces one character

  • [] Matches one of more characters in the brackets

  • /data/sales/**/*.csv Gets all csv files under /data/sales

  • /data/sales/20??/**/ Gets all files in the 20th century

  • /data/sales/*/*/*.csv Gets csv files two levels under /data/sales

  • /data/sales/2004/*/12/[XY]1?.csv Gets all csv files in 2004 in December starting with X or Y prefixed by a two-digit number

Partition Root Path: If you have partitioned folders in your file source with a key=value format (for example, year=2019), then you can assign the top level of that partition folder tree to a column name in your data flow data stream.

First, set a wildcard to include all paths that are the partitioned folders plus the leaf files that you wish to read.

Screenshot of partition source file settings in mapping data flow source transformation.

Use the Partition Root Path setting to define what the top level of the folder structure is. When you view the contents of your data via a data preview, you'll see that ADF will add the resolved partitions found in each of your folder levels.

Partition root path

List of files: This is a file set. Create a text file that includes a list of relative path files to process. Point to this text file.

Column to store file name: Store the name of the source file in a column in your data. Enter a new column name here to store the file name string.

After completion: Choose to do nothing with the source file after the data flow runs, delete the source file, or move the source file. The paths for the move are relative.

To move source files to another location post-processing, first select "Move" for file operation. Then, set the "from" directory. If you're not using any wildcards for your path, then the "from" setting will be the same folder as your source folder.

If you have a source path with wildcard, your syntax will look like this below:

/data/sales/20??/**/*.csv

You can specify "from" as

/data/sales

And "to" as

/backup/priorSales

In this case, all files that were sourced under /data/sales are moved to /backup/priorSales.

Note

File operations run only when you start the data flow from a pipeline run (a pipeline debug or execution run) that uses the Execute Data Flow activity in a pipeline. File operations do not run in Data Flow debug mode.

Filter by last modified: You can filter which files you process by specifying a date range of when they were last modified. All date-times are in UTC.

Enable change data capture: If true, you will get new or changed files only from the last run. Initial load of full snapshot data will always be gotten in the first run, followed by capturing new or changed files only in next runs. For more details, see Change data capture.

Screenshot showing Enable change data capture.

Sink properties

In the sink transformation, you can write to either a container or folder in Azure Data Lake Storage Gen2. the Settings tab lets you manage how the files get written.

sink options

Clear the folder: Determines whether or not the destination folder gets cleared before the data is written.

File name option: Determines how the destination files are named in the destination folder. The file name options are:

  • Default: Allow Spark to name files based on PART defaults.
  • Pattern: Enter a pattern that enumerates your output files per partition. For example, loans[n].csv will create loans1.csv, loans2.csv, and so on.
  • Per partition: Enter one file name per partition.
  • As data in column: Set the output file to the value of a column. The path is relative to the dataset container, not the destination folder. If you have a folder path in your dataset, it will be overridden.
  • Output to a single file: Combine the partitioned output files into a single named file. The path is relative to the dataset folder. Please be aware that the merge operation can possibly fail based upon node size. This option is not recommended for large datasets.

Quote all: Determines whether to enclose all values in quotes

umask

You can optionally set the umask for files using POSIX read, write, execute flags for owner, user and group.

Pre-processing and post-processing commands

You can optionally execute Hadoop filesystem commands before or after writing to an ADLS Gen2 sink. The following commands are supported:

  • cp
  • mv
  • rm
  • mkdir

Examples:

  • mkdir /folder1
  • mkdir -p folder1
  • mv /folder1/*.* /folder2/
  • cp /folder1/file1.txt /folder2
  • rm -r /folder1

Parameters are also supported through expression builder, for example:

mkdir -p {$tempPath}/commands/c1/c2 mv {$tempPath}/commands/*.* {$tempPath}/commands/c1/c2

By default, folders are created as user/root. Refer to the top level container with '/'.

Lookup activity properties

To learn details about the properties, check Lookup activity.

GetMetadata activity properties

To learn details about the properties, check GetMetadata activity

Delete activity properties

To learn details about the properties, check Delete activity

Legacy models

Note

The following models are still supported as-is for backward compatibility. You are suggested to use the new model mentioned in above sections going forward, and the ADF authoring UI has switched to generating the new model.

Legacy dataset model

Property Description Required
type The type property of the dataset must be set to AzureBlobFSFile. Yes
folderPath Path to the folder in Data Lake Storage Gen2. If not specified, it points to the root.

Wildcard filter is supported. Allowed wildcards are * (matches zero or more characters) and ? (matches zero or single character). Use ^ to escape if your actual folder name has a wildcard or this escape char is inside.

Examples: filesystem/folder/. See more examples in Folder and file filter examples.
No
fileName Name or wildcard filter for the files under the specified "folderPath". If you don't specify a value for this property, the dataset points to all files in the folder.

For filter, the wildcards allowed are * (matches zero or more characters) and ? (matches zero or single character).
- Example 1: "fileName": "*.csv"
- Example 2: "fileName": "???20180427.txt"
Use ^ to escape if your actual file name has a wildcard or this escape char is inside.

When fileName isn't specified for an output dataset and preserveHierarchy isn't specified in the activity sink, the copy activity automatically generates the file name with the following pattern: "Data.[activity run ID GUID].[GUID if FlattenHierarchy].[format if configured].[compression if configured]", for example, "Data.0a405f8a-93ff-4c6f-b3be-f69616f1df7a.txt.gz". If you copy from a tabular source using a table name instead of a query, the name pattern is "[table name].[format].[compression if configured]", for example, "MyTable.csv".
No
modifiedDatetimeStart Files filter based on the attribute Last Modified. The files are selected if their last modified time is greater than or equal to modifiedDatetimeStart and less than modifiedDatetimeEnd. The time is applied to the UTC time zone in the format of "2018-12-01T05:00:00Z".

The overall performance of data movement is affected by enabling this setting when you want to do file filter with huge amounts of files.

The properties can be NULL, which means no file attribute filter is applied to the dataset. When modifiedDatetimeStart has a datetime value but modifiedDatetimeEnd is NULL, it means the files whose last modified attribute is greater than or equal to the datetime value are selected. When modifiedDatetimeEnd has a datetime value but modifiedDatetimeStart is NULL, it means the files whose last modified attribute is less than the datetime value are selected.
No
modifiedDatetimeEnd Files filter based on the attribute Last Modified. The files are selected if their last modified time is greater than or equal to modifiedDatetimeStart and less than modifiedDatetimeEnd. The time is applied to the UTC time zone in the format of "2018-12-01T05:00:00Z".

The overall performance of data movement is affected by enabling this setting when you want to do file filter with huge amounts of files.

The properties can be NULL, which means no file attribute filter is applied to the dataset. When modifiedDatetimeStart has a datetime value but modifiedDatetimeEnd is NULL, it means the files whose last modified attribute is greater than or equal to the datetime value are selected. When modifiedDatetimeEnd has a datetime value but modifiedDatetimeStart is NULL, it means the files whose last modified attribute is less than the datetime value are selected.
No
format If you want to copy files as is between file-based stores (binary copy), skip the format section in both the input and output dataset definitions.

If you want to parse or generate files with a specific format, the following file format types are supported: TextFormat, JsonFormat, AvroFormat, OrcFormat, and ParquetFormat. Set the type property under format to one of these values. For more information, see the Text format, JSON format, Avro format, ORC format, and Parquet format sections.
No (only for binary copy scenario)
compression Specify the type and level of compression for the data. For more information, see Supported file formats and compression codecs.
Supported types are **GZip**, **Deflate**, **BZip2**, and **ZipDeflate**.
Supported levels are Optimal and Fastest.
No

Tip

To copy all files under a folder, specify folderPath only.
To copy a single file with a given name, specify folderPath with a folder part and fileName with a file name.
To copy a subset of files under a folder, specify folderPath with a folder part and fileName with a wildcard filter.

Example:

{
    "name": "ADLSGen2Dataset",
    "properties": {
        "type": "AzureBlobFSFile",
        "linkedServiceName": {
            "referenceName": "<Azure Data Lake Storage Gen2 linked service name>",
            "type": "LinkedServiceReference"
        },
        "typeProperties": {
            "folderPath": "myfilesystem/myfolder",
            "fileName": "*",
            "modifiedDatetimeStart": "2018-12-01T05:00:00Z",
            "modifiedDatetimeEnd": "2018-12-01T06:00:00Z",
            "format": {
                "type": "TextFormat",
                "columnDelimiter": ",",
                "rowDelimiter": "\n"
            },
            "compression": {
                "type": "GZip",
                "level": "Optimal"
            }
        }
    }
}

Legacy copy activity source model

Property Description Required
type The type property of the copy activity source must be set to AzureBlobFSSource. Yes
recursive Indicates whether the data is read recursively from the subfolders or only from the specified folder. When recursive is set to true and the sink is a file-based store, an empty folder or subfolder isn't copied or created at the sink.
Allowed values are true (default) and false.
No
maxConcurrentConnections The upper limit of concurrent connections established to the data store during the activity run. Specify a value only when you want to limit concurrent connections. No

Example:

"activities":[
    {
        "name": "CopyFromADLSGen2",
        "type": "Copy",
        "inputs": [
            {
                "referenceName": "<ADLS Gen2 input dataset name>",
                "type": "DatasetReference"
            }
        ],
        "outputs": [
            {
                "referenceName": "<output dataset name>",
                "type": "DatasetReference"
            }
        ],
        "typeProperties": {
            "source": {
                "type": "AzureBlobFSSource",
                "recursive": true
            },
            "sink": {
                "type": "<sink type>"
            }
        }
    }
]

Legacy copy activity sink model

Property Description Required
type The type property of the copy activity sink must be set to AzureBlobFSSink. Yes
copyBehavior Defines the copy behavior when the source is files from a file-based data store.

Allowed values are:
- PreserveHierarchy (default): Preserves the file hierarchy in the target folder. The relative path of the source file to the source folder is identical to the relative path of the target file to the target folder.
- FlattenHierarchy: All files from the source folder are in the first level of the target folder. The target files have autogenerated names.
- MergeFiles: Merges all files from the source folder to one file. If the file name is specified, the merged file name is the specified name. Otherwise, it's an autogenerated file name.
No
maxConcurrentConnections The upper limit of concurrent connections established to the data store during the activity run. Specify a value only when you want to limit concurrent connections. No

Example:

"activities":[
    {
        "name": "CopyToADLSGen2",
        "type": "Copy",
        "inputs": [
            {
                "referenceName": "<input dataset name>",
                "type": "DatasetReference"
            }
        ],
        "outputs": [
            {
                "referenceName": "<ADLS Gen2 output dataset name>",
                "type": "DatasetReference"
            }
        ],
        "typeProperties": {
            "source": {
                "type": "<source type>"
            },
            "sink": {
                "type": "AzureBlobFSSink",
                "copyBehavior": "PreserveHierarchy"
            }
        }
    }
]

Change data capture

Azure Data Factory can get new or changed files only from Azure Data Lake Storage Gen2 by enabling Enable change data capture in the mapping data flow source transformation. With this connector option, you can read new or updated files only and apply transformations before loading transformed data into destination datasets of your choice.

Make sure you keep the pipeline and activity name unchanged, so that the checkpoint can always be recorded from the last run to get changes from there. If you change your pipeline name or activity name, the checkpoint will be reset, and you will start from the beginning in the next run.

When you debug the pipeline, the Enable change data capture works as well. Be aware that the checkpoint will be reset when you refresh your browser during the debug run. After you are satisfied with the result from debug run, you can publish and trigger the pipeline. It will always start from the beginning regardless of the previous checkpoint recorded by debug run.

In the monitoring section, you always have the chance to rerun a pipeline. When you are doing so, the changes are always gotten from the checkpoint record in your selected pipeline run.

For a list of data stores supported as sources and sinks by the copy activity, see Supported data stores.