[Bug report] Gravitino REST Server Defaults to 'jdbc' Instead of Using Catalog Name from Spark Configuration

[dataageek]

Version

main branch

Describe what's wrong

I have multiple catalogs created using the Iceberg JDBC catalog. When I attempt to consume these catalogs via the Gravitino REST server (using the REST catalog through Spark), the implementation always defaults to using jdbc as the catalog name instead of the catalog name specified in the Spark configuration.

For example, when running the following command:

./spark-sql \
    --conf spark.sql.catalog.iceberg_jdbc_catalog=org.apache.iceberg.spark.SparkCatalog \
    --conf spark.sql.catalog.iceberg_jdbc_catalog.warehouse="/mnt/c/iceberg-warehouse" \
    --conf spark.sql.catalog.iceberg_jdbc_catalog.type=rest \
    --conf spark.sql.catalog.iceberg_jdbc_catalog.uri=http://127.0.0.1:9991/iceberg \
    --conf spark.sql.extensions=org.apache.iceberg.spark.extensions.IcebergSparkSessionExtensions

I expect Spark to use the iceberg_jdbc_catalog catalog, as specified in the configuration. However, the Gravitino REST server implementation instead looks for a catalog named jdbc.

Although I can override this behavior using the catalog-backend-name property in gravitino.conf, I would prefer a dynamic mechanism that fetches the catalog name from the Spark configuration directly.

Error message and/or stacktrace

No error message. just not working as expected

How to reproduce

0.7.0

Additional context

No response

[dataageek]

hi @FANNG1 Could you please let me know if there is a workaround for this issue in Spark?

[FANNG1]

You mean the Iceberg catalog name in Iceberg REST server side is decided by the Spark client ? It seems a little odd, The catalog backend name is initiated when startup.