JingsongLi commented on code in PR #11015:
URL: https://github.com/apache/paimon/pull/11015#discussion_r4523018688
##########
paimon-flink/paimon-flink-common/src/main/java/org/apache/paimon/flink/sink/FlinkSinkBuilder.java:
##########
@@ -351,6 +353,61 @@ protected DataStreamSink<?> buildDynamicBucketSink(
protected DataStreamSink<?> buildForFixedBucket(DataStream<InternalRow>
input) {
int bucketNums = table.bucketSpec().getNumBuckets();
+ Options options = Options.fromMap(table.options());
+ // sink.striped-bucket-write.enabled is the sole switch;
writers-per-bucket only applies
+ // when it is
Review Comment:
**[P2] Prevent stripes from publishing duplicate primary-key vector indexes**
I reproduced this on `f58ed0ff8` with a fixed one-bucket `FIRST_ROW` table,
a physically compacted source file, and a vector index enabled afterwards.
Deletion vectors are disabled. I used the real channel computer to route two
keys to separate write-only writers, then prepared and committed both outputs.
Each writer built an ANN segment over the same compacted source. FAST search
returned no candidates because the search planner rejects competing segments at
the same level. With one writer and the same rows, one segment was published
and FAST found the compacted row. Could we reject striped writes with this
index maintenance enabled, or give each bucket exactly one index publisher?
--
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]