SteNicholas commented on code in PR #7669:
URL: https://github.com/apache/hudi/pull/7669#discussion_r1090269465
##########
hudi-client/hudi-flink-client/src/main/java/org/apache/hudi/table/action/commit/FlinkDeletePartitionCommitActionExecutor.java:
##########
@@ -98,4 +103,43 @@ private List<String> getAllExistingFileIds(String
partitionPath) {
// because new commit is not complete. it is safe to mark all existing
file Ids as old files
return
table.getSliceView().getLatestFileSlices(partitionPath).map(FileSlice::getFileId).distinct().collect(Collectors.toList());
}
+
+ /**
+ * Check if there are any pending table service actions (requested +
inflight) on a table affecting the partitions to
+ * be dropped.
+ * <p>
+ * This check is to prevent a drop-partition from proceeding should a
partition have a table service action in
+ * the pending stage. If this is allowed to happen, the filegroup that is an
input for a table service action, might
+ * also be a candidate for being replaced. As such, when the table service
action and drop-partition commits are
+ * committed, there will be two commits replacing a single filegroup.
+ * <p>
+ * For example, a timeline might have an execution order as such:
+ * 000.replacecommit.requested (clustering filegroup_1 + filegroup_2 ->
filegroup_3)
+ * 001.replacecommit.requested, 001.replacecommit.inflight,
0001.replacecommit (drop_partition to replace filegroup_1)
+ * 000.replacecommit.inflight (clustering is executed now)
+ * 000.replacecommit (clustering completed)
+ * For an execution order as shown above, 000.replacecommit and
001.replacecommit will both flag filegroup_1 to be replaced.
+ * This will cause downstream duplicate key errors when a map is being
constructed.
+ */
+ private void checkPreconditions() {
Review Comment:
Could this introduce an untility class into util package of
hudi-client-common? And then the Flink and Spark client could invoke this
method.
--
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]
