manager: cgronlun

[!INCLUDE[appliesto-ss-xxxx-xxxx-xxx-md-winonly](../includes/appliesto-ss-xxxx-xxxx-xxx-md-winonly.md)]

**Applies to:** SQL Server 2016 R Services, SQL Server 2017 Machine Learning Services (R and Python)

SQL Server 2016 R Services is the first release of SQL Server to include integrated R support. SQL Server 2016 Service Pack 1 (SP1) includes several major improvements, including the ability to run external scripts. If you are a SQL Server 2016 customer, you should consider installing SP1 or later.

, @script = N'

As a last resort, you can open files on the server to determine the installed version. To do so, locate the rlauncher.config file to get the location of the R runtime and the current working directory. We recommend that you make and open a copy of the file so that you don't accidentally change any properties.

`C:\Program Files\Microsoft SQL Server\MSSQL14.<instance_name>\R_SERVICES\bin\x64\RGui.exe`

The R console displays the version information on startup. For example, the following version represents the default configuration for SQL Server 2017 CTP 2.0:

*Microsoft R Open 3.3.3*

*The enhanced R distribution from Microsoft*

*Microsoft packages Copyright (C) 2017 Microsoft*

2. Get the value for **PYTHONHOME**.

3. Get the value of the current working directory.

Check to see whether more than one copy of the R libraries is installed on the computer. This duplication can happen if:

* You install Microsoft R Client in addition to SQL Server.

* A different set of R libraries was installed by using R Tools for Visual Studio, R Studio, Microsoft R Client, or another R IDE.

* The computer hosts multiple instances of SQL Server, and more than one instance uses machine learning.

The errors that you see when you attempt to run R code can come from any of the following sources:

When you work with the service for the first time, it can be difficult to tell which messages originate from which services. We recommend that you capture not only the exact message text, but the context in which you saw the message. Note the client software that you're using to run machine learning code:

`C:\Program Files\Microsoft SQL Server\MSSQL14.SQL2016\MSSQL\Log\ExtensibilityLog`

To remove R or Python problems from consideration, you can run this script, which starts the R or Python runtime and passes data back and forth.

`C:\Program Files\Microsoft SQL Server\MSSQL14.<instance_name>\MSSQL\Log\ExtensibilityLog `

For example, the following log messages are related to the extensibility framework:

This might indicate that the worker accounts that run external scripts cannot access the instance.

This message might mean that your security settings are preventing setup from creating the pool of worker accounts that are needed to run external scripts.

* *Satellite Session Manager initialization failed*

2. Open the ExtLaunchErrorlog file, and look for the string *ErrorCode*. Review the message that's associated with the ErrorCode.

* *The SQL Server Launchpad (MSSQLSERVER) service failed to start due to the following error: <text>*

2. Open the bootstrap log subfolder that is specific to extensibility.

3. If you need to submit a support request, add the entire contents of this folder to a zipped file. For example, C:\Program Files\Microsoft SQL Server\130\Setup Bootstrap\Log\LOG\ExtensibilityLog.

3. To enable script execution, create roles or add users to the following roles, as necessary:

- All but *db_owner*: Require EXECUTE ANY EXTERNAL SCRIPT.

4. Note whether you changed any default startup accounts when you installed SQL Server 2016.

2. Determine whether a firewall rule has been created for SQL Server. For security reasons, in a default installation, it might not be possible for remote R or Python client to connect to the instance. For more information, see [Troubleshooting connecting to SQL Server](../database-engine/configure-windows/troubleshoot-connecting-to-the-sql-server-database-engine.md).

[Troubleshoot machine learning in SQL Server](machine-learning-troubleshooting-faq.md)

+ Use feature selection to remove unnecessary columns.

+ If you are using an open source algorithm, consider a similar implementation using the corresponding algorithm in MicrosoftML or RevoScaleR. These packages have been optimized for deployment scenarios.

+ After the model has been rationalized and the size reduced using the preceding steps, see if the [memCompress](https://www.rdocumentation.org/packages/base/versions/3.4.1/topics/memCompress) function in base R can be used to reduce the size of the model before passing it to SQL Server. This option is best when the model is close to the 2 GB limit.

After you have enabled FileTable, to write the model, you get a path from SQL using the FileTable API, and then write the model to that location from your code. When you need to read the model, you get the path from SQL and then call the model using the path from your script. For more information, see [Access FileTables with File Input-Output APIs](../relational-databases/blob/access-filetables-with-file-input-output-apis.md).

+ If your code is relatively simple, you can embed it in a T-SQL user-defined function without modification, as described in these samples:

+ If the code is more complex, use the R package **sqlrutils** to convert your code. This package is designed to help experienced R users write good stored procedure code.

Package repositories are useful in these scenarios:

- **Easier offline installation**: To install package to an offline server requires that you also download all package dependencies, Using miniCRAN makes it easier to get all dependencies in the correct format.

makeRepo(pkgs_expanded, path = local_repo, repos = CRAN_mirror, type = "win.binary", Rversion = "3.3");

```

At this point you should have a folder containing the packages you needed, and any additional packages that were required. The path should be similar to this example: C:\mylocalrepo\bin\windows\contrib\3.3 and it should contain a collection of zipped packages. Do not unzip the packages or rename any files.

EXECUTE sp_execute_external_script

+ [Introduction to memory-optimized tables](https://docs.microsoft.com/sql/relational-databases/in-memory-oltp/introduction-to-memory-optimized-tables)

Provides specific tips on how to optimize the external script to avoid known problems. It is most useful to **data scientists**.

> [!NOTE]

> While much of the information in this section applies to R in general, some information is specific to RevoScaleR analytic functions. Detailed performance guidance is not available for **revoscalepy** and other supported Python libraries.

>

Summarizes what data was used the two case studies, how performance was tested, and how the optimizations affected results.