Offsets committed out of order with Spark DataSource API V2 Hive Streaming Sink

问题

I am using sink to save Spark(2.3) Structured Streaming DataFrame into Hive table with a custom sink implementation.

The code is as follows.

val df = spark.readStream.format("socket").option("host", "localhost").option("port", 19191).load().as[String]


val query = df.map { s => val records = s.split(",") assert(records.length >= 4)
        (records(0).toInt, records(1), records(2), records(3))
     }


query.selectExpr("_1 as eid", "_2 as name", "_3 as salary", "_4 as designation").
      writeStream.
      format("hive-streaming").
      option("metastore", ".....").
      option("db", "test").
      option("table", "test_employee").
      option("checkpointLocation", "/checkpoints/employee/checkpoint").
      queryName("socket-hive-streaming").
      start()

This would lead to following runtime error.

ERROR streaming.MicroBatchExecution: Query socket-hive-streaming [id =  ......, runId = ......] terminated with error
java.lang.RuntimeException: Offsets committed out of order: 1 followed by 0
        at scala.sys.package$.error(package.scala:27)
        at org.apache.spark.sql.execution.streaming.TextSocketSource.commit(socket.scala:146)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch$1$$anonfun$apply$mcV$sp$4.apply(MicroBatchExecution.scala:356)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch$1$$anonfun$apply$mcV$sp$4.apply(MicroBatchExecution.scala:355)
        at scala.collection.Iterator$class.foreach(Iterator.scala:893)
        at scala.collection.AbstractIterator.foreach(Iterator.scala:1336)
        at scala.collection.IterableLike$class.foreach(IterableLike.scala:72)
        at org.apache.spark.sql.execution.streaming.StreamProgress.foreach(StreamProgress.scala:25)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch$1.apply$mcV$sp(MicroBatchExecution.scala:355)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch$1.apply(MicroBatchExecution.scala:338)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch$1.apply(MicroBatchExecution.scala:338)
        at org.apache.spark.sql.execution.streaming.ProgressReporter$class.reportTimeTaken(ProgressReporter.scala:271)
        at org.apache.spark.sql.execution.streaming.StreamExecution.reportTimeTaken(StreamExecution.scala:58)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution.org$apache$spark$sql$execution$streaming$MicroBatchExecution$$constructNextBatch(MicroBatchExecution.scala:338)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$runActivatedStream$1$$anonfun$apply$mcZ$sp$1.apply$mcV$sp(MicroBatchExecution.scala:128)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$runActivatedStream$1$$anonfun$apply$mcZ$sp$1.apply(MicroBatchExecution.scala:121)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$runActivatedStream$1$$anonfun$apply$mcZ$sp$1.apply(MicroBatchExecution.scala:121)
        at org.apache.spark.sql.execution.streaming.ProgressReporter$class.reportTimeTaken(ProgressReporter.scala:271)
        at org.apache.spark.sql.execution.streaming.StreamExecution.reportTimeTaken(StreamExecution.scala:58)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution$$anonfun$runActivatedStream$1.apply$mcZ$sp(MicroBatchExecution.scala:121)
        at org.apache.spark.sql.execution.streaming.ProcessingTimeExecutor.execute(TriggerExecutor.scala:56)
        at org.apache.spark.sql.execution.streaming.MicroBatchExecution.runActivatedStream(MicroBatchExecution.scala:117)
        at org.apache.spark.sql.execution.streaming.StreamExecution.org$apache$spark$sql$execution$streaming$StreamExecution$$runStream(StreamExecution.scala:279)
        at org.apache.spark.sql.execution.streaming.StreamExecution$$anon$1.run(StreamExecution.scala:189)

回答1:

There are two ways to solve your problem:

1. Delete/clear your checkpoint: /checkpoints/employee/checkpoint on your machine

2. Using the other source that maintains offset, like Kafka

The reason you meet this issue is that the socket doesn't maintain offset information.

When you restart your job which receives input data from socket 9999, the first thing that your job does is trying to recover the state from /checkpoints/employee/checkpoint, and it finds that your current offset that has been recorded is 1. Then you input other messages to socket 9999, your job finds that the offset from socket 9999 is 0. So it throws this exception.

来源：https://stackoverflow.com/questions/50493552/offsets-committed-out-of-order-with-spark-datasource-api-v2-hive-streaming-sink