From 99b14a78e38b081efbdda59243626c244335ea7c Mon Sep 17 00:00:00 2001
From: zhangminglei <18717838093@126.com>
Date: Fri, 21 May 2021 21:59:47 +0800
Subject: [PATCH] [HUDI-1918] Fix incorrect keyBy field cause serious data
 skew, to avoid multiple subtasks write to a partition at the same time
 (#2972)

---
 .../java/org/apache/hudi/streamer/HoodieFlinkStreamer.java    | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/hudi-flink/src/main/java/org/apache/hudi/streamer/HoodieFlinkStreamer.java b/hudi-flink/src/main/java/org/apache/hudi/streamer/HoodieFlinkStreamer.java
index 88ab2b658..bd485634d 100644
--- a/hudi-flink/src/main/java/org/apache/hudi/streamer/HoodieFlinkStreamer.java
+++ b/hudi-flink/src/main/java/org/apache/hudi/streamer/HoodieFlinkStreamer.java
@@ -88,8 +88,8 @@ public class HoodieFlinkStreamer {
         .name("kafka_source")
         .uid("uid_kafka_source")
         .map(new RowDataToHoodieFunction<>(rowType, conf), TypeInformation.of(HoodieRecord.class))
-        // Key-by partition path, to avoid multiple subtasks write to a partition at the same time
-        .keyBy(HoodieRecord::getPartitionPath)
+        // Key-by record key, to avoid multiple subtasks write to a partition at the same time
+        .keyBy(HoodieRecord::getRecordKey)
         .transform(
             "bucket_assigner",
             TypeInformation.of(HoodieRecord.class),