[SPARK-20894] Error while checkpointing to HDFS

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.1
• Fix Version/s: 2.3.0
• Component/s: Structured Streaming

Environment

Ubuntu, Spark 2.1.1, hadoop 2.7

• Assignee: Shixiong Zhu
• Reporter: kant kodali
• Created: 26/May/17 06:03
• Updated: 11/Jun/19 02:19
• Resolved: 08/Aug/17 03:56

Description

Dataset<Row> df2 = df1.groupBy(functions.window(df1.col("Timestamp5"), "24 hours", "24 hours"), df1.col("AppName")).count();

StreamingQuery query = df2.writeStream().foreach(new KafkaSink()).option("checkpointLocation","/usr/local/hadoop/checkpoint").outputMode("update").start();
query.awaitTermination();

This for some reason fails with the Error

ERROR Executor: Exception in task 0.0 in stage 1.0 (TID 1)
java.lang.IllegalStateException: Error reading delta file /usr/local/hadoop/checkpoint/state/0/0/1.delta of HDFSStateStoreProvider[id = (op=0, part=0), dir = /usr/local/hadoop/checkpoint/state/0/0]: /usr/local/hadoop/checkpoint/state/0/0/1.delta does not exist

I did clear all the checkpoint data in /usr/local/hadoop/checkpoint/ and all consumer offsets in Kafka from all brokers prior to running and yet this error still persists.

Issue Links

• links to: [Github] Pull Request #18149 (zsxwing)
• links to: [Github] Pull Request #18179 (zsxwing)