Skip to content

sql-server-linux-bulk-operations.md

Latest commit

 

History

History
203 lines (146 loc) · 7.76 KB

sql-server-linux-bulk-operations.md

File metadata and controls

203 lines (146 loc) · 7.76 KB
title Configure Bulk Import Operations for SQL Server on Linux
description Learn how to configure and use the bulkadmin server role or the ADMINISTER BULK OPERATIONS permission for bulk data import in SQL Server on Linux.
author VanMSFT
ms.author vanto
ms.reviewer matripathy, randolphwest
ms.date 03/11/2026
ms.service sql
ms.subservice linux
ms.topic how-to
ms.custom
linux-related-content
ai-usage ai-assisted
monikerRange =sql-server-ver17 || =sql-server-linux-ver17

Configure bulk import operations for SQL Server on Linux (preview)

[!INCLUDE sqlserver2025-linux]

Important

This feature is currently in preview.

Starting with [!INCLUDE sssql25-md] Cumulative Update (CU) 3, you can use the bulkadmin server role or the ADMINISTER BULK OPERATIONS permission to perform bulk data import operations on [!INCLUDE ssnoversion-md] running on Linux. Previously, only members of the sysadmin server role could run BULK INSERT or OPENROWSET(BULK...) on Linux.

[!INCLUDE ssnoversion-md] on Linux enforces additional file system and path validation checks for bulk operations, beyond what's required on Windows. An administrator must:

  • Grant appropriate [!INCLUDE ssnoversion-md] permissions to the user
  • Grant Linux file system permissions on the data files
  • Explicitly approve directory paths using mssql-conf

Prerequisites

  • [!INCLUDE sssql25-md] CU 3 or later version on Linux
  • Administrative access to the Linux host
  • Administrative access to the [!INCLUDE ssnoversion-md] instance

Configure the Linux file system

Before you can run bulk import operations, the mssql service account must have read access to the data files on the Linux file system.

  1. Create a directory for your bulk data files:

    mkdir -p /tmp/bulkload/sales/

  2. Create a sample data file:

    cat > /tmp/bulkload/sales/loadsalesdata.csv << EOF
Id,CustomerName,OrderDate,Amount
1,John Doe,2026-02-01,500.75
2,Jane Smith,2026-02-05,1500.20
3,Mark Lee,2026-02-10,320.50
4,Alice Johnson,2026-02-15,785.00
5,Bob Brown,2026-02-20,930.40
EOF

  3. Grant read permission to the [!INCLUDE ssnoversion-md] process account (mssql) on the data file:

    setfacl -m u:mssql:r /tmp/bulkload/sales/loadsalesdata.csv

Configure allowed paths with mssql-conf

An administrator must approve the directories from which bulk operations can read, using the bulkadmin.allowedpathslist setting in mssql-conf. This change takes effect immediately and doesn't require a [!INCLUDE ssnoversion-md] service restart.

sudo /opt/mssql/bin/mssql-conf set bulkadmin.allowedpathslist "/tmp/bulkload/sales"

To specify multiple directories, separate each path with a colon (:):

sudo /opt/mssql/bin/mssql-conf set bulkadmin.allowedpathslist "/tmp/bulkload/sales:/tmp/bulkload/marketing"

Alternatively, you can specify a parent directory to allow all subdirectories under it.

Path restrictions

The following restrictions apply to paths configured for bulk operations:

  • The path must be an absolute path. Relative paths containing . or .. aren't allowed.
  • The root path (/) isn't allowed.
  • The path must not exceed 4,096 characters.
  • The path must not contain invalid characters (null, newline, carriage return, or tab).
  • Symbolic links aren't allowed.
  • The path must refer to a directory.

Forbidden paths

The system blocks some critical paths by design. You can't use these locations as source data file paths for bulk operations, even if you add them to the allowed paths list:

  • /var/opt/mssql (or the [!INCLUDE ssnoversion-md] installation directory)
  • /var/opt/azcmagent/certs
  • /var/opt/azcmagent/tokens

Bulk operations that read source data from these paths fail for security reasons. This restriction doesn't apply to the ERRORFILE output path, which uses the [!INCLUDE ssnoversion-md] data directory by default.

Set up SQL Server permissions

After you configure the Linux file system and allowed paths, set up the [!INCLUDE ssnoversion-md] permissions for the user who performs the bulk import.

  1. Connect to the [!INCLUDE ssnoversion-md] instance and create a sample database and table:

    CREATE DATABASE demodbforbulkinsert;
GO

USE demodbforbulkinsert;
GO

CREATE TABLE dbo.Sales (
    Id INT NOT NULL PRIMARY KEY,
    CustomerName NVARCHAR(200) NOT NULL,
    OrderDate DATE NOT NULL,
    Amount DECIMAL(18, 2) NOT NULL
);
GO

  2. Create a login for the bulk import user:

    USE master;
GO

CREATE LOGIN BulkLoadUser WITH PASSWORD = '<strong_password>';
GO

  3. Grant the ADMINISTER BULK OPERATIONS permission or add the login to the bulkadmin server role. Use one of the following options:

    -- Option 1: Add to the bulkadmin server role
ALTER SERVER ROLE bulkadmin ADD MEMBER BulkLoadUser;
GO
    -- Option 2: Grant the permission directly
GRANT ADMINISTER BULK OPERATIONS TO BulkLoadUser;
GO

  4. Create a database user and grant the necessary table permissions:

    USE demodbforbulkinsert;
GO

CREATE USER BulkLoadUser FOR LOGIN BulkLoadUser;
GO

GRANT INSERT, SELECT ON dbo.Sales TO BulkLoadUser;
GO

Run a bulk import

Connect to the database as the BulkLoadUser login and run the bulk import:

USE demodbforbulkinsert;
GO

BULK INSERT dbo.Sales
FROM '/tmp/bulkload/sales/loadsalesdata.csv'
WITH (
    FIRSTROW = 2,
    FIELDTERMINATOR = ',',
    ERRORFILE = '/var/opt/mssql/data/bulk_errors'
);
GO

-- Verify the imported data
SELECT * FROM dbo.Sales;
GO

The same permissions, Linux file system configuration, and path approval steps apply to INSERT ... SELECT * FROM OPENROWSET(BULK...) statements. For more information, see Import bulk data by using BULK INSERT or OPENROWSET(BULK...).

Upgrade and downgrade behavior

Starting with [!INCLUDE sssql25-md] CU 3, [!INCLUDE ssnoversion-md] on Linux supports bulk operations using the bulkadmin role or ADMINISTER BULK OPERATIONS permission.

If you uninstall CU 3 or downgrade to an earlier cumulative update:

  • Previously granted permissions remain assigned but aren't functional.
  • Bulk operations require sysadmin permissions, as in versions prior to CU 3.

Related content