hanyuzheng7 commented on code in PR #29624:
URL: https://github.com/apache/flink/pull/29624#discussion_r4385438746
##########
flink-table/flink-table-planner/src/main/java/org/apache/flink/table/planner/plan/nodes/exec/stream/StreamExecValues.java:
##########
@@ -35,12 +35,21 @@
import java.util.List;
/** Stream {@link ExecNode} that read records from given values. */
+// Version 1: Initial version
@ExecNodeMetadata(
name = "stream-exec-values",
version = 1,
producedTransformations = CommonExecValues.VALUES_TRANSFORMATION,
minPlanVersion = FlinkVersion.v1_15,
minStateVersion = FlinkVersion.v1_15)
+// Version 2: VALUES runs as a FLIP-27 source that checkpoints the position of
the next row to
Review Comment:
I added version 2 here because the new FLIP-27 source can be chained with
the sink writer (or an async operator), and then that operator gets max
parallelism 1 from VALUES. When I tested it, restoring an old checkpoint failed
with "Max parallelism mismatch", even with uids. So old compiled plans (version
1) still use the legacy source. I used v2_5 because this will first be released
in 2.5. FLINK-38738 did the same with v2_3. If this doesn't make it into 2.5, I
will change it to the next version and regenerate the restore test plan.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
