Skip to main content
Version: Beta 🚧

Java Gateway Process Exited Before Sending its Port Number

Issue​

When running an EMR-connected Jupyter notebook, you see the following error when running the Tecton SDK function get_historical_features():

~/.local/lib/python3.8/site-packages/pyspark/java_gateway.py in launch_gateway(conf, popen_kwargs)
   107             if not os.path.isfile(conn_info_file):
--> 108                 raise Exception("Java gateway process exited before sending its port number")
   109
Exception: Java gateway process exited before sending its port number
The above exception was the direct cause of the following exception:

[... Python stack trace ...]

Note that other Tecton SDK functions may continue to work properly, such as get_data_source() and list_feature_views().

Resolution​

The most likely issue is that you are connected to a Python kernel instead of a PySpark kernel. When you create a notebook in EMR, you have the option to use a number of kernels, and you should choose PySpark. If you have already created the notebook, you can click in the upper-right-hand corner and select PySpark instead of Python.

get_historical_features() explicitly uses Spark to execute, however, some other Tecton functions only make calls to Tecton APIs to retrieve data, which does not require a Spark context.

Was this page helpful?

Happy React is loading...