SQLFlow releases several Docker images that contains the SQLFlow server, MySQL server, sample datasets, Jupyter Notebook server, and the SQLFlow plugin for Jupyter.

You can use these Docker images for either local trying out or production deployment.

1. Install Docker Community Edition on your PC/Macbook/Server.

2. Pull the latest SQLFlow Docker images. Or you can also build the Docker image from source code following this guide.

   docker pull sqlflow/sqlflow
   docker pull sqlflow/sqlflow:mysql
   docker pull sqlflow/sqlflow:jupyter
   

1. Type the below command to start three containers to start a MySQL server, SQLFlow server and a Jupyter notebook server.

   docker run --name=sqlflow-mysql -d -p 8888:8888 sqlflow/sqlflow:mysql
   docker run --net=container:sqlflow-mysql -d sqlflow/sqlflow:latest sqlflowserver
   docker run --net=container:sqlflow-mysql -d sqlflow/sqlflow:jupyter
   

2. You can also use a specified version (e.g. v0.4.0) of the SQLFlow server by changing the second line above to docker run --net=container:sqlflow-mysql -d sqlflow/sqlflow:v0.4.0 sqlflowserver.

3. Open a web browser, go to localhost:8888, open any tutorial notebook like iris-dnn.ipynb file, then you can follow the tutorial and run the SQL statements to run the training and prediction.

If you have your own database setup, below steps enables running a separated SQLFlow server and Jupyter Notebook which connects to your own database service.

For MySQL, please refer to run_with_mysql.

For Hive, please refer to run_with_hive.

For MaxCompute, please refer to run_with_maxcompute.