Spark Job failed on YARN - - apache-spark

Question

I am trying to execute the Spark job in YARN Cluster using the following configurations.
/usr/bin/spark-submit
--class com.example.DriverClass
--master yarn-cluster
app.jar
hdfs:///user/spark/file1.parquet
hdfs:///user/spark/file2.parquet
hdfs:///user/spark/output
20151217052915
--num-executors 20
--executor-memory 12288M
--executor-cores 5
--driver-memory 6G
--conf spark.yarn.executor.memoryOverhead=1332
We are executing with 20 executors and each executor we are passing as 12 GB memory for this job.
Do we have to increase the size of spark.yarn.executor.memoryOverhead property ?
Error log:
15/12/18 15:47:39 WARN scheduler.TaskSetManager: Lost task 2.0 in stage 5.0 (TID 117, lpdn0185.com): java.lang.OutOfMemoryError: GC overhead limit exceeded
at org.apache.spark.util.collection.ExternalAppendOnlyMap$ExternalIterator$$anonfun$next$1.apply(ExternalAppendOnlyMap.scala:336)
at org.apache.spark.util.collection.ExternalAppendOnlyMap$ExternalIterator$$anonfun$next$1.apply(ExternalAppendOnlyMap.scala:331)
at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59)
at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47)
at org.apache.spark.util.collection.ExternalAppendOnlyMap$ExternalIterator.next(ExternalAppendOnlyMap.scala:331)
at org.apache.spark.util.collection.ExternalAppendOnlyMap$ExternalIterator.next(ExternalAppendOnlyMap.scala:227)
at org.apache.spark.InterruptibleIterator.next(InterruptibleIterator.scala:43)
at scala.collection.Iterator$class.foreach(Iterator.scala:727)
at org.apache.spark.InterruptibleIterator.foreach(InterruptibleIterator.scala:28)
at org.apache.spark.rdd.SubtractedRDD.integrate$1(SubtractedRDD.scala:110)
at org.apache.spark.rdd.SubtractedRDD.compute(SubtractedRDD.scala:119)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.UnionRDD.compute(UnionRDD.scala:87)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:61)
at org.apache.spark.scheduler.Task.run(Task.scala:64)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:203)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
15/12/18 15:47:39 INFO scheduler.TaskSetManager: Starting task 2.1 in stage 5.0 (TID 119, lpdn0185.com, PROCESS_LOCAL, 4237 bytes)
15/12/18 15:47:39 WARN scheduler.TaskSetManager: Lost task 3.0 in stage 5.0 (TID 118, lpdn0185.com): FetchFailed(BlockManagerId(2, lpdn0185..com, 37626), shuffleId=4, mapId=42, reduceId=3, message=
org.apache.spark.shuffle.FetchFailedException: Error in opening FileSegmentManagedBuffer{file=/hdfs1/yarn/nm/usercache/phdpentcustcdibtch/appcache/application_1449986083135_60217/blockmgr-34a2e882-6b36-42c6-bcff-03d9bc5ef80b/0c/shuffle_4_42_0.data, offset=5899394, length=46751}
at org.apache.spark.shuffle.hash.BlockStoreShuffleFetcher$.org$apache$spark$shuffle$hash$BlockStoreShuffleFetcher$$unpackBlock$1(BlockStoreShuffleFetcher.scala:67)
at org.apache.spark.shuffle.hash.BlockStoreShuffleFetcher$$anonfun$3.apply(BlockStoreShuffleFetcher.scala:83)
at org.apache.spark.shuffle.hash.BlockStoreShuffleFetcher$$anonfun$3.apply(BlockStoreShuffleFetcher.scala:83)
at scala.collection.Iterator$$anon$13.hasNext(Iterator.scala:371)
at org.apache.spark.util.CompletionIterator.hasNext(CompletionIterator.scala:32)
at org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala:39)
at org.apache.spark.Aggregator.combineCombinersByKey(Aggregator.scala:91)
at org.apache.spark.shuffle.hash.HashShuffleReader.read(HashShuffleReader.scala:44)
at org.apache.spark.rdd.ShuffledRDD.compute(ShuffledRDD.scala:92)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.SubtractedRDD.integrate$1(SubtractedRDD.scala:110)
at org.apache.spark.rdd.SubtractedRDD.compute(SubtractedRDD.scala:119)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.UnionRDD.compute(UnionRDD.scala:87)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:35)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:277)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:244)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:61)
at org.apache.spark.scheduler.Task.run(Task.scala:64)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:203)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.IOException: Error in opening FileSegmentManagedBuffer{file=/hdfs1/yarn/nm/usercache/user1/appcache/application_1449986083135_60217/blockmgr-34a2e882-6b36-42c6-bcff-03d9bc5ef80b/0c/shuffle_4_42_0.data, offset=5899394, length=46751}
at org.apache.spark.network.buffer.FileSegmentManagedBuffer.createInputStream(FileSegmentManagedBuffer.java:113)
at org.apache.spark.storage.ShuffleBlockFetcherIterator$$anonfun$3.apply(ShuffleBlockFetcherIterator.scala:300)
at org.apache.spark.storage.ShuffleBlockFetcherIterator$$anonfun$3.apply(ShuffleBlockFetcherIterator.scala:300)
at scala.util.Try$.apply(Try.scala:161)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.next(ShuffleBlockFetcherIterator.scala:300)
at org.apache.spark.storage.ShuffleBlockFetcherIterator.next(ShuffleBlockFetcherIterator.scala:53)
... 30 more
Caused by: java.io.FileNotFoundException: /hdfs1/yarn/nm/usercache/user1/appcache/application_1449986083135_60217/blockmgr-34a2e882-6b36-42c6-bcff-03d9bc5ef80b/0c/shuffle_4_42_0.data (No such file or directory)
at java.io.FileInputStream.open0(Native Method)
at java.io.FileInputStream.open(FileInputStream.java:195)
at java.io.FileInputStream.<init>(FileInputStream.java:138)
at org.apache.spark.network.buffer.FileSegmentManagedBuffer.createInputStream(FileSegmentManagedBuffer.java:98)
... 35 more
)
Appreciate your help on this.

Answer 1

I had the same issue for about several weeks. Exactly speaking, every time I got
slightly different errors including what you got. Basically, in my case, I think, compared to cluster capability, data was too big.
In brief, what I tried was
increased executor memory
increased spark.yarn.executor.memoryOverhead upto 20% of executorMemory (10% is default with minimum of 384)
checked your build version and spark version
increased or reduced number of executors depending on the number of cluster nodes (how many executors are allocated per node?)
optimized codes
-minimize shuffling
e.g. avoid groupByKey, replaced by reduceByKey, aggregateByKey, or combineByKey
-minimize temporary files internally cached
e.g. optimized transforms / number of transforms
considered the number of partitions
(how many partitioned parquet files?)
in my case, repartitioning via coalesce or partitioner, etc. didn't work, actually, the performance got worse performance when repartitioning
Hope this works!