How to deploy an app on SQL Server big data cluster (preview)

THIS TOPIC APPLIES TO:yesSQL Server (starting with 2019)noAzure SQL DatabasenoAzure SQL Data Warehouse noParallel Data Warehouse

This article describes how to deploy and manage R and Python script as an application inside a SQL Server 2019 big data cluster (preview).

What's new and improved

  • A single command-line utility to manage cluster and app.
  • Simplified app deployment while providing granular control through spec files.
  • Support hosting additional application types - SSIS and MLeap (new in CTP 2.3)
  • VS Code Extension to manage application deployment

Applications are deployed and managed using mssqlctl command-line utility. This article provides examples of how to deploy apps from the command line. To learn how to use this in Visual Studio Code refer to VS Code Extension.

The following types of apps are supported:

  • R and Python apps (functions, models and apps)
  • MLeap Serving
  • SQL Server Integration Services (SSIS)



In SQL Server 2019 (preview) CTP 3.0 you can create, delete, describe, initialize, list run and update your application. The following table describes the application deployment commands that you can use with mssqlctl.

Command Description
mssqlctl login Sign into a SQL Server big data cluster
mssqlctl app create Create application.
mssqlctl app delete Delete application.
mssqlctl app describe Describe application.
mssqlctl app init Kickstart new application skeleton.
mssqlctl app list List application(s).
mssqlctl app run Run application.
mssqlctl app update Update application.

You can get help with the --help parameter as in the following example:

mssqlctl app create --help

The following sections describe these commands in more detail.

Sign in

Before you deploy or interact with applications, first sign in to your SQL Server big data cluster with the mssqlctl login command. Specify the external IP address of the controller-svc-external service (for example: https://ip-address:30080) along with the user name and password to the cluster.

mssqlctl login --controller-endpoint https://<ip-address-of-controller-svc-external>:30080 --controller-username <user-name>


If you are using AKS, you need to run the following command to get the IP address of the mgmtproxy-svc-external service by running this command in a bash or cmd window:

kubectl get svc mgmtproxy-svc-external -n <name of your big data cluster>

Kubeadm or Minikube

If you are using Kubeadm or Minikube run the following command to get the IP address to login in to the cluster

kubectl get node --selector=''

Create an app

To create an application, you use mssqlctl with the app create command. These files reside locally on the machine that you are creating the app from.

Use the following syntax to create a new app in big data cluster:

mssqlctl app create --spec <directory containing spec file>

The following command shows an example of what this command might look like:

mssqlctl app create --spec ./addpy

This assumes that you have your application stored in the addpy folder. This folder should also contain a specification file for the application, called called spec.yaml. Please see the Application Deployment page for more information on the spec.yaml file.

To deploy this app sample app, create the following files in a directory called addpy:

  • Copy the following Python code into this file:
    def add(x,y):
          result = x+y
          return result
  • spec.yaml. Copy the following code into this file:
    name: add-app #name of your python script
    version: v1  #version of the app
    runtime: Python #the language this app uses (R or Python)
    src: ./ #full path to the location of the app
    entrypoint: add #the function that will be called upon execution
    replicas: 1  #number of replicas needed
    poolsize: 1  #the pool size that you need your app to scale
    inputs:  #input parameters that the app expects and the type
       x: int
       y: int
    output: #output parameter the app expects and the type
       result: int

Then, run the command below:

mssqlctl app create --spec ./addpy

You can check if the app is deployed using the list command:

mssqlctl app list

If the deployment is not complete you should see the state show WaitingforCreate as the following example:

    "name": "add-app",
    "state": "WaitingforCreate",
    "version": "v1"

After the deployment is successful, you should see the state change to Ready status:

    "name": "add-app",
    "state": "Ready",
    "version": "v1"

List an app

You can list any apps that were successfully created with the app list command.

The following command lists all available applications in your big data cluster:

mssqlctl app list

If you specify a name and version, it lists that specific app and its state (Creating or Ready):

mssqlctl app list --name <app_name> --version <app_version>

The following example demonstrates this command:

mssqlctl app list --name add-app --version v1

You should see output similar to the following example:

    "name": "add-app",
    "state": "Ready",
    "version": "v1"

Run an app

If the app is in a Ready state, you can use it by running it with your specified input parameters. Use the following syntax to run an app:

mssqlctl app run --name <app_name> --version <app_version> --inputs <inputs_params>

The following example command demonstrates the run command:

mssqlctl app run --name add-app --version v1 --inputs x=1,y=2

If the run was successful, you should see your output as specified when you created the app. The following is an example.

  "changedFiles": [],
  "consoleOutput": "",
  "errorMessage": "",
  "outputFiles": {},
  "outputParameters": {
    "result": 3
  "success": true

Create an app skeleton

The init command provides a scaffold with the relevant artifacts that is required for deploying an app. The example below creates hello you can do this by running the following command.

mssqlctl app init --name hello --version v1 --template python

This will create a folder called hello. You can cd into the directory and inspect the generated files in the folder. spec.yaml defines the app, such as name, version and source code. You can edit the spec to change name, version, input and outputs.

Here is a sample output from the init command that you will see in the folder

Describe an app

The describe command provides detailed information about the app including the end point in your cluster. This is typically used by an app developer to build an app using the swagger client and using the webservice to interact with the app in a RESTful manner. See Consume applications on big data clusters for more information.

  "input_param_defs": [
      "name": "x",
      "type": "int"
      "name": "y",
      "type": "int"
  "links": {
    "app": "",
    "swagger": ""
  "name": "add-app",
  "output_param_defs": [
      "name": "result",
      "type": "int"
  "state": "Ready",
  "version": "v1"

Delete an app

To delete an app from your big data cluster, use the following syntax:

mssqlctl app delete --name add-app --version v1

Next steps

Explore how to integrate apps deployed on SQL Server big data clusters in your own applications at Consume applications on big data clusters for more information. You can also check out additional samples at App Deploy Samples.

For more information about SQL Server big data clusters, see What are SQL Server 2019 big data clusters?.