Customer found in Azure Databricks that when two notebooks connected to the same DataBricks cluster, TempView cannot be created in one of the notebooks. Issue is resolved when they restart the cluster.
Testing was done at H2O. The first notebook to connect to H2O is able to access the TempView tables (including ones created by second notebook).
Produced an error in Python and Scala:
Create one notebook, run:
No error occurs
Create a second notebook and connect to same cluster, Repeat the script above changing TempView name:
AnalysisException: 'Table or view note found: output_df_2; line 1 pos 38
It will error saying TempView not found. This is because the first notebook has the H2O cluster pointing to its SQL context and all TempView tables are being saved to the first notebooks SQL context.
Recommendation: Separate the caching of data for the two notebooks (although they connect to the same H2O cluster)