tianfy created FLINK-39840:
------------------------------
Summary: Oracle pipeline connector emits incomplete table
identifiers for data and schema change events
Key: FLINK-39840
URL: https://issues.apache.org/jira/browse/FLINK-39840
Project: Flink
Issue Type: Bug
Components: Flink CDC
Reporter: tianfy
The Oracle pipeline connector should preserve the full Oracle table identifier
when emitting pipeline events.
Currently some paths convert Debezium table IDs to Flink CDC table IDs by
dropping the database/catalog part and only keeping schema.table. Data change
records, schema change records, and CreateTableEvent can therefore use
different table identifiers for the same Oracle table.
This can break downstream routing and schema evolution in jobs that rely on
database.schema.table identifiers.
Expected behavior:
* data change events use database.schema.table
* schema change events use database.schema.table
* CreateTableEvent uses the same database.schema.table identifier
The fix should use the table identifier from the SourceRecord source struct
where possible instead of parsing the Kafka topic name, and should keep the
catalog/database part when converting Debezium TableId to Flink CDC TableId.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
