TheR1sing3un commented on code in PR #12451:
URL: https://github.com/apache/hudi/pull/12451#discussion_r1881573645
##########
hudi-client/hudi-client-common/src/main/java/org/apache/hudi/table/ConsistentHashingBucketInsertPartitioner.java:
##########
@@ -29,5 +29,14 @@ public interface ConsistentHashingBucketInsertPartitioner {
* @param partition partition to set Consistent Hashing nodes
* @param nodes nodes from clustering plan
*/
+ /**
+ * Set pending consistent hashing for partition, only used in executing
clustering
+ * When call this method, the bulk insert will directly use the pending
metadata as the consistent hash metadata for writing data to after-resizing
buckets.
+ * Used in the case of executing bulk insert.
+ * NOTE: This method should be called before the bulk insert operation, and
will skip building identifiers from records, just use the pending metadata.
+ * For which not calling this method, the bulk insert will use the committed
metadata as the bucket metadata and disallow writing data to the
pending-resizing buckets.
+ * @param partition partition to set Consistent Hashing nodes
+ * @param nodes nodes from clustering plan
+ */
Review Comment:
> Then shall we move the instantiation of the metadata once in the
constructor? Not sure if there are some SE/DE issue.
What you means is that we remove the method:
`ConsistentHashBucketInsertPartitioner::addHashingChildrenNodes`? And then in
the clustering scenario, the partitioner is instantiated directly by carrying
these metadata as construct parameters?
--
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]
