[HUDI-4931] Explore fat jar option for gcs-connector lib used during GCS Ingestion

• Type: Task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: deltastreamer
• Epic Link: Implement DeltaStreamer Source for cloud object stores
• Assignee: Unassigned
• Reporter: Pramod Biligiri
• Created: 27/Sep/22 10:50
• Updated: 10/Mar/23 01:02

Description

Currently, the GCS Ingestion (HUDI-4850) expects recent versions of Jars like protobuf and Guava to be provided to spark-submit explicitly, to override older versions shipped with Spark. These Jars are used by the gcs-connector which is a library from Google that helps connect to GCS. For more details see https://docs.google.com/document/d/1VfvtdvhXw6oEHPgZ_4Be2rkPxIzE0kBCNUiVDsXnSAA/edit# (section titled "Configure Spark to use newer versions of some Jars").

See if it's possible to create a shaded+fat jar of gcs-connector for this use case instead, and avoid specifying things to spark-submit on the command line.

An alternate approach to consider for the long term is HUDI-4930 (slim bundles).

Issue Links

• is a child of: HUDI-4850 Implement DeltaStreamer Source for Google Cloud Storage