[BUG] Spark Job Progress widget not working after %%configure -f
rankol opened this issue · comments
Describe the bug
I am using a PySpark kernel inside a Jupyter notebook attached to an EMR cluster. When I first start my PySpark session and run some code to calculate Pi I am able to see a widget that indicates the progress of the submitted code on the EMR cluster.
When I attempt to use this command to re-configure the Spark session I am no long able to see the widget properly.
%%configure -f {"name":"remotesparkmagics", "executorMemory":"34490M", "driverMemory":"34490M", "conf": {"spark.pyspark.python": "python3", "spark.pyspark.virtualenv.enabled": "true", "spark.pyspark.virtualenv.type": "native", "spark.pyspark.virtualenv.bin.path": "/usr/bin/virtualenv"}}
The error message is
There was an error connecting to your cluster: JSONDecodeError('Expecting value: line 1 column 1 (char 0)')
This is most often caused by an attempt to monitor a session that is no longer active.
I have been able to workaround this by running %%cleanup -f
then something such as sc.list_packages()
to start a new session. After that I can run my code to calculate Pi and I can see the Spark Job Progress widget working properly.
To Reproduce
- Start Spark session
- Run code to calculate Pi using Spark
- Re-configure Spark session with
%%configure -f
- Run code to calculate Pi
- Run
%%cleanup -f
- Run
sc.list_packages()
to start new Spark session - Run code to calculate Pi
Expected behavior
I would like to determine a way to have the widget work without needing to cleanup my old session if that is possible.
Screenshots
Currently can't upload my screenshots.
Versions:
- SparkMagic (unsure)
- Livy 0.7.1
- Spark 3.1.2