Επεξεργασία

Κοινή χρήση μέσω


Install SQL Server 2022 Machine Learning Services (Python and R) on Linux

Applies to: SQL Server 2022 (16.x) - Linux

This article guides you in the installation of SQL Server Machine Learning Services on Linux. Python and R scripts can be executed in-database using Machine Learning Services.

You can install Machine Learning Services on Ubuntu and Red Hat Enterprise Linux (RHEL). Currently, SUSE Linux Enterprise Server (SLES) is unsupported.

You can install ML Services on a Docker container running a Linux distribution. Inside the Docker container, the steps would be the same as below.

For more information, see the Supported platforms section in the installation guidance for SQL Server on Linux.

Important

This article refers to SQL Server 2022 (16.x). For SQL Server 2019 on Linux, see to Install SQL Server 2019 Machine Learning Services (Python and R) on Linux. For SQL Server on Windows, see Install SQL Server 2022 Machine Learning Services (Python and R) on Windows.

Pre-install checklist

  • Install SQL Server on Linux and verify the installation.

  • Check the SQL Server Linux repositories for the Python and R extensions. If you already configured source repositories for the database engine install, you can run the mssql-server-extensibility package install commands using the same repo registration.

  • You should have a tool for running T-SQL commands.

    • You can use Azure Data Studio, a free database tool that runs on Linux, Windows, and macOS.
  • Restarting the SQL Server instance during this installation process will be required.

Package list

On an internet-connected device, packages are downloaded and installed independently of the database engine using the package installer for each operating system.

Available installation packages for SQL Server 2022 (16.x) on Linux:

Package name Applies-to Description
mssql-server-extensibility All Extensibility framework used to run Python and R.

Install mssql-server-extensibility package

  1. Configure repositories for installing and upgrading SQL Server on Linux corresponding to the Linux distribution. Install the SQL Server extensibility feature with the package mssql-server-extensibility and associated dependency libssl-dev.

    Ubuntu

    sudo apt-get install mssql-server-extensibility libssl-dev
    

    RHEL

    yum install mssql-server-extensibility
    
  2. Review and accept the End User License Agreement (EULA) for SQL Server ML Services.

    sudo /opt/mssql/bin/mssql-conf set EULA accepteulaml Y
    

    To complete acceptance of the EULA, the SQL Server instance must be restarted.

    sudo systemctl restart mssql-server
    

Install runtimes and packages

Setup R support

Install R runtime

  1. Download and install the version of R that is desired. Choose a version of R 4.2 or higher, available for download directly from cran.r-project.org. Follow the instructions for the desired runtime.

  2. Open an admin R terminal:

    sudo R
    
  3. Install dependencies for CompatibilityAPI and RevoScaleR by running the following:

    # R Terminal
    install.packages("iterators", lib="/usr/lib/R/library")
    install.packages("foreach", lib="/usr/lib/R/library")
    install.packages("R6", lib="/usr/lib/R/library")
    install.packages("jsonlite", lib="/usr/lib/R/library")
    
  4. Download and install CompatibilityAPI and RevoScaleR for Linux.

    install.packages("https://aka.ms/sqlml/r4.2/linux/CompatibilityAPI_1.1.0_R_x86_64-pc-linux-gnu.tar.gz", repos=NULL, lib="/usr/lib/R/library")
    install.packages("https://aka.ms/sqlml/r4.2/linux/RevoScaleR_10.0.1_R_x86_64-pc-linux-gnu.tar.gz", repos=NULL, lib="/usr/lib/R/library")
    
  5. Verify RevoScaleR installation from the R terminal.

    library("RevoScaleR")
    

Configure R runtime with SQL Server

  1. Configure the installed R runtime with SQL Server for Linux, where path/to/ is the file path to the R binary, and RFolderVersion is the version-specific folder name for your installation of R runtime, for example, R4.2.

    sudo /opt/mssql/bin/mssql-conf set extensibility rbinpath /usr/lib/R/bin/R
    sudo /opt/mssql/bin/mssql-conf set extensibility datadirectories /usr/lib/R
    
  2. Restart the Launchpadd service.

    systemctl restart mssql-launchpadd.service
    
  3. Configure SQL Server for Linux to allow external scripts using the sp_configure system stored procedure.

    EXECUTE sp_configure 'external scripts enabled', 1;
    GO
    
    RECONFIGURE;
    GO
    
  4. Verify the installation by executing a simple T-SQL command to return the version of R:

    EXECUTE sp_execute_external_script
        @script = N'print(R.version)',
        @language = N'R';
    GO
    

Setup Python support

Install Python runtime

  1. Download and install Python. Choose version Python 3.10 specifically, available for download directly from python.org. Follow the instructions provided. Also, install the shared python runtime library for the runtime version. For example, to install libpython3.10 for Ubuntu: sudo apt-get install libpython3.10.

  2. Download and install revoscalepy for the root user.

    sudo pip install dill numpy==1.22.0 pandas patsy python-dateutil
    sudo pip install https://aka.ms/sqlml/python3.10/linux/revoscalepy-10.0.1-py3-none-any.whl --target=/usr/lib/python3.10/dist-packages
    
  3. Verify the revoscalepy installation from the Python terminal. Verify the library can be imported.

    import revoscalepy
    

Configure Python runtime with SQL Server

  1. Configure the installed Python runtime with SQL Server, where pythonbinbath is set to the path of the installed python binary, and datadirectories includes the path where the packages are installed for the desired version of python, for example, /usr/lib/python3.10/dist-packages. Use the following script with your actual installation path:

    sudo /opt/mssql/bin/mssql-conf set extensibility pythonbinpath /usr/bin/python3.10
    sudo /opt/mssql/bin/mssql-conf set extensibility datadirectories /usr/lib:/usr/lib/python3.10/dist-packages
    
  2. Restart the Launchpadd service.

    systemctl restart mssql-launchpadd.service
    
  3. Configure SQL Server for Linux to allow external scripts using the sp_configure system stored procedure.

    EXECUTE sp_configure 'external scripts enabled', 1;
    GO
    
    RECONFIGURE;
    GO
    
  4. Verify the installation by executing a simple T-SQL command to return the version of python:

    EXECUTE sp_execute_external_script
        @script = N'import sys;print(sys.version)',
        @language = N'Python';
    GO
    

Install Java

To install the Java language extension, see Install SQL Server Java Language Extension on Linux.

Verify installation

To validate installation, use any of the following methods:

  • Run a T-SQL script that executes a system stored procedure invoking Python or R using a query tool.

  • Execute the following SQL command to test R execution in SQL Server. Errors? Try a service restart, sudo systemctl restart mssql-server.service.

    EXECUTE sp_execute_external_script
        @language = N'R',
        @script = N'
          OutputDataSet <- InputDataSet', @input_data_1 = N'SELECT 1 AS hello'
        WITH RESULT SETS
    (
            ([hello] INT NOT NULL)
    );
    GO
    
  • Execute the following SQL command to test Python execution in SQL Server.

    EXECUTE sp_execute_external_script
        @language = N'Python',
        @script = N'
          OutputDataSet = InputDataSet;
          ', @input_data_1 = N'SELECT 1 AS hello'
        WITH RESULT SETS
    (
            ([hello] INT NOT NULL)
    );
    GO
    

Offline installation

Follow the Offline installation instructions for steps on installing the packages. Find your download site, and then download specific packages using the package list below.

Tip

Several of the package management tools provide commands that can help you determine package dependencies. For yum, use sudo yum deplist [package]. For Ubuntu, use sudo apt-get install --reinstall --download-only [package name] followed by dpkg -I [package name].deb.

Standalone RevoScale packages for Python and R runtime

RevoScale packages are also supported as a standalone package with Python and R runtimes. In order to setup Python or R runtime for the standalone scenario, follow the instructions in the Install Python runtime and Install R runtime sections respectively.