Xiaobing Fang created FLINK-41387:
-------------------------------------
Summary: Reduce unnecessary ADD COLUMN requests in the Fluss
pipeline sink
Key: FLINK-41387
URL: https://issues.apache.org/jira/browse/FLINK-41387
Project: Flink
Issue Type: Improvement
Components: Flink CDC
Reporter: Xiaobing Fang
h3. Motivation
The goal is to reduce unnecessary ADD COLUMN requests in Fluss-to-Fluss
pipelines where columns are added to the target before the source. When the
source emits the corresponding AddColumnEvent, the sink submits all requested
columns to Fluss. If any already exist, Fluss rejects the request, and only
then does the sink check the target schema. Across many tables, this produces a
failed DDL request for each such event.
h3. Expected behavior
Check the target schema before issuing ADD COLUMN and submit only missing
columns. For example, if an event requests columns a and b and the target
already has a with the matching type, send one request to add b. If both
columns already exist with matching types, complete the event without issuing
an ADD COLUMN request.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
