Upgrade from spark 2.4 to spark 3.3 in aws for ML pre processing fail with stackoverflow error

[amada2023]

Hi,

I use a glue script in aws to do some pre-processing for ML pipeline. Have been using glue 2.0 with spark 2.4. Now I am trying to move it to latest glue 4.0 which uses spark 3.3.
when running script on latest verion it fails with error
Error Category: UNCLASSIFIED_ERROR; Py4JJavaError: An error occurred while calling o6517.count. java.lang.StackOverflowError
same error on using df.cache()
dataframe is loaded post reading from parquet files path and then cached.

with spark 2.4 everything was working fine but with spark 3.3 caching is throwing these stackoverflow errors. Is there any configuration change in sparkcontext with 3.3 ?

Thanks
Aashima