How to reset spark structured streaming data to last available offset

Question

I am running a structured stream application with Kafka. I found that if for some reason the system is down for days... The checkpoint become stale and offsets corresponding to the checkpoint is not found in Kafka. How do I let Spark Structured Streaming app pick the last available offset and start from there. I tried setting offset reset to earlier/latest but the system was crashing with following error:

org.apache.kafka.clients.consumer.OffsetOutOfRangeException: Offsets out of range with no configured reset policy for partitions: {MyTopic-574=6559828}
at org.apache.kafka.clients.consumer.internals.Fetcher.parseCompletedFetch(Fetcher.java:970)
at org.apache.kafka.clients.consumer.internals.Fetcher.fetchedRecords(Fetcher.java:490)
at org.apache.kafka.clients.consumer.KafkaConsumer.pollForFetches(KafkaConsumer.java:1259)
at org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1187)
at org.apache.kafka.clients.consumer.KafkaConsumer.poll(KafkaConsumer.java:1115)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer.fetchData(KafkaDataConsumer.scala:470)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer.org$apache$spark$sql$kafka010$InternalKafkaConsumer$$fetchRecord(KafkaDataConsumer.scala:361)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer$$anonfun$get$1.apply(KafkaDataConsumer.scala:251)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer$$anonfun$get$1.apply(KafkaDataConsumer.scala:234)
at org.apache.spark.util.UninterruptibleThread.runUninterruptibly(UninterruptibleThread.scala:77)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer.runUninterruptiblyIfPossible(KafkaDataConsumer.scala:209)
at org.apache.spark.sql.kafka010.InternalKafkaConsumer.get(KafkaDataConsumer.scala:234)
at org.apache.spark.sql.kafka010.KafkaDataConsumer$class.get(KafkaDataConsumer.scala:64)
at org.apache.spark.sql.kafka010.KafkaDataConsumer$CachedKafkaDataConsumer.get(KafkaDataConsumer.scala:500)
at org.apache.spark.sql.kafka010.KafkaMicroBatchInputPartitionReader.next(KafkaMicroBatchReader.scala:357)
at org.apache.spark.sql.execution.datasources.v2.DataSourceRDD$$anon$1.hasNext(DataSourceRDD.scala:49)
at org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala:37)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage1.processNext(Unknown Source)
at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$13$$anon$1.hasNext(WholeStageCodegenExec.scala:636)
at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:409)
at org.apache.spark.sql.catalyst.expressions.GeneratedClass$GeneratedIteratorForCodegenStage2.processNext(Unknown Source)
at org.apache.spark.sql.execution.BufferedRowIterator.hasNext(BufferedRowIterator.java:43)
at org.apache.spark.sql.execution.WholeStageCodegenExec$$anonfun$13$$anon$1.hasNext(WholeStageCodegenExec.scala:636)
at org.apache.spark.sql.execution.UnsafeExternalRowSorter.sort(UnsafeExternalRowSorter.java:216)
at org.apache.spark.sql.execution.SortExec$$anonfun$1.apply(SortExec.scala:108)
at org.apache.spark.sql.execution.SortExec$$anonfun$1.apply(SortExec.scala:101)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$24.apply(RDD.scala:836)
at org.apache.spark.rdd.RDD$$anonfun$mapPartitionsInternal$1$$anonfun$apply$24.apply(RDD.scala:836)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:52)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:324)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:288)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:90)
at org.apache.spark.scheduler.Task.run(Task.scala:123)
at org.apache.spark.executor.Executor$TaskRunner$$anonfun$10.apply(Executor.scala:408)
at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1360)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:414)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)

Can you show the code of your consumer together with your configurations? Also, which versions are you using? — mike
I had same issue & I have followed this approach - As you mentioned offsets in your checkpoint is too old, can you please take backup & delete your checkpoint location & Try to start your spark application. That should work.. — Srinivas
@Srinivas Thanks so much for your response. We want to not touch the real time streaming pipeline... If offsets are old we want Spark to ignore the offsets and continue. I have set failOnDataLoss to false but that did not solve my issue. If above issue happen I have to clean the checkpoint i.e user data then restart Spark. This is not acceptable for production system. — Alchemist

Giorgos Myrianthous Giorgos Myrianthous · Accepted Answer · 2020-04-18T16:28:58

If the system was down for days then it means that some logs might have been compacted. More precisely, your application attempts to read offset 6559828 from the 574th partition in topic MyTopic.

In order to find the earliest available offset per partition, you can simply run the following command:

bin/kafka-run-class.sh kafka.tools.GetOffsetShell \
    --broker-list localhost:9092 \
    --topic MyTopic \
    --time -2

How to reset spark structured streaming data to last available offset

1 Answers