Wrong FS on loading json with spark from s3 - hadoop

I'm trying to load geojson file with spark and magellan library
My code for loading is:
val polygons = spark.read.format("magellan").option("type", "geojson").load(inJson)
Where inJson is path to my json on s3:
s3n://bucket-name/geojsons/file.json
Error with stack trace:
0.3 in stage 0.0 (TID 3, ip-172-31-19-102.eu-west-1.compute.internal, executor 1): java.lang.IllegalArgumentException: Wrong FS:
s3n://bucket-name/geojsons/file.json, expected:
hdfs://ip-172-31-27-182.eu-west-1.compute.internal:8020 at
org.apache.hadoop.fs.FileSystem.checkPath(FileSystem.java:653) at
org.apache.hadoop.hdfs.DistributedFileSystem.getPathName(DistributedFileSystem.java:194)
at
org.apache.hadoop.hdfs.DistributedFileSystem.access$000(DistributedFileSystem.java:106)
at
org.apache.hadoop.hdfs.DistributedFileSystem$3.doCall(DistributedFileSystem.java:304)
at
org.apache.hadoop.hdfs.DistributedFileSystem$3.doCall(DistributedFileSystem.java:299)
at
org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81)
at
org.apache.hadoop.hdfs.DistributedFileSystem.open(DistributedFileSystem.java:312)
at org.apache.hadoop.fs.FileSystem.open(FileSystem.java:773) at
magellan.mapreduce.WholeFileReader.nextKeyValue(WholeFileReader.scala:45)
at
org.apache.spark.rdd.NewHadoopRDD$$anon$1.hasNext(NewHadoopRDD.scala:199)
at
org.apache.spark.InterruptibleIterator.hasNext(InterruptibleIterator.scala:39)
at scala.collection.Iterator$$anon$12.hasNext(Iterator.scala:439) at
scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at
scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:408) at
scala.collection.Iterator$class.foreach(Iterator.scala:893) at
scala.collection.AbstractIterator.foreach(Iterator.scala:1336) at
scala.collection.TraversableOnce$class.foldLeft(TraversableOnce.scala:157)
at scala.collection.AbstractIterator.foldLeft(Iterator.scala:1336)
at
scala.collection.TraversableOnce$class.fold(TraversableOnce.scala:212)
at scala.collection.AbstractIterator.fold(Iterator.scala:1336) at
org.apache.spark.rdd.RDD$$anonfun$fold$1$$anonfun$20.apply(RDD.scala:1086)
at
org.apache.spark.rdd.RDD$$anonfun$fold$1$$anonfun$20.apply(RDD.scala:1086)
at
org.apache.spark.SparkContext$$anonfun$33.apply(SparkContext.scala:1980)
at
org.apache.spark.SparkContext$$anonfun$33.apply(SparkContext.scala:1980)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:99) at
org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:282)
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)
The problem occurs only when I run it on more than one machine, so it works fine on EMR cluster with master and 1 instance in core group, but fails like that with 10 instances in core group

This turned out to be problem within Magellan WholeFileReader. It was getting default FileSystem.
It was solved with this pull request
The solutions was like so:
- val fs = FileSystem.get(conf)
+ val fs = path.getFileSystem(conf)

If you are running on EMR, you can just use "s3://bucket/path" instead of "s3n://...."

Related

Pyspark streaming write in elasticsearch

Is there a way to write through Spark Streaming, reading from kafka and write in ElasticSearch?
I tried something like... As explained in the ElasticSearch docs (with very little documentation regarding pyspark):
sc = SparkContext("local[2]", appName="TwitterStreamKafka")
ssc = StreamingContext(sc, batchIntervalSeconds)
topic = url_topic
tweets = KafkaUtils.createStream(ssc, zkQuorum, "spark-streaming-consumer", {topic: 1})
tweets.pprint()
conf = {"es.resource": "credentials/credential"} # assume Elasticsearch is running on localhost defaults
if tweets.count() > 0:
tweets.foreachRDD(lambda rdd: rdd.saveAsNewAPIHadoopFile(
path='-',
outputFormatClass="org.elasticsearch.hadoop.mr.EsOutputFormat",
keyClass="org.apache.hadoop.io.NullWritable",
valueClass="org.elasticsearch.hadoop.mr.LinkedMapWritable",
conf=conf))
ssc.start()
ssc.awaitTermination()
But it doesn't work. The error is:
17/11/10 17:16:35 ERROR Utils: Aborting task
org.elasticsearch.hadoop.rest.EsHadoopInvalidRequest: Found unrecoverable error [127.0.0.1:9200] returned Bad Request(400) - failed to parse; Bailing out..
at org.elasticsearch.hadoop.rest.RestClient.processBulkResponse(RestClient.java:251)
at org.elasticsearch.hadoop.rest.RestClient.bulk(RestClient.java:203)
at org.elasticsearch.hadoop.rest.RestRepository.tryFlush(RestRepository.java:222)
at org.elasticsearch.hadoop.rest.RestRepository.flush(RestRepository.java:244)
at org.elasticsearch.hadoop.rest.RestRepository.close(RestRepository.java:269)
at org.elasticsearch.hadoop.mr.EsOutputFormat$EsRecordWriter.doClose(EsOutputFormat.java:214)
at org.elasticsearch.hadoop.mr.EsOutputFormat$EsRecordWriter.close(EsOutputFormat.java:196)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$4.apply(SparkHadoopMapReduceWriter.scala:155)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$4.apply(SparkHadoopMapReduceWriter.scala:144)
at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1375)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$.org$apache$spark$internal$io$SparkHadoopMapReduceWriter$$executeTask(SparkHadoopMapReduceWriter.scala:159)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$3.apply(SparkHadoopMapReduceWriter.scala:89)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$3.apply(SparkHadoopMapReduceWriter.scala:88)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:335)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
17/11/10 17:16:35 ERROR SparkHadoopMapReduceWriter: Task attempt_20171110171633_0003_r_000000_0 aborted.
17/11/10 17:16:35 ERROR Executor: Exception in task 0.0 in stage 3.0 (TID 3)
org.apache.spark.SparkException: Task failed while writing rows
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$.org$apache$spark$internal$io$SparkHadoopMapReduceWriter$$executeTask(SparkHadoopMapReduceWriter.scala:178)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$3.apply(SparkHadoopMapReduceWriter.scala:89)
at org.apache.spark.internal.io.SparkHadoopMapReduceWriter$$anonfun$3.apply(SparkHadoopMapReduceWriter.scala:88)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87)
at org.apache.spark.scheduler.Task.run(Task.scala:108)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:335)
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)
This is the command I use to execute it:
spark-submit --jars elasticsearch-hadoop-5.6.4.jar,spark-streaming-kafka-0-10-assembly_2.11-2.2.0.jar es_spark_write.py
I am using spark 2.2.0
The messages from Kafka are key-message json like these:
(u'urls', u'{"token": "secret_token", "count": 2443}')

IllegalDataException from DateUtil.java when saving spark streaming dataframe to phoenix

I am using kafka + spark streaming to stream messages and do analytics, then saving to phoenix. Some spark job fail several times per day with the following error message:
org.apache.phoenix.schema.IllegalDataException:
java.lang.IllegalArgumentException: Invalid format: ""
at org.apache.phoenix.util.DateUtil$ISODateFormatParser.parseDateTime(DateUtil.java:297)
at org.apache.phoenix.util.DateUtil.parseDateTime(DateUtil.java:163)
at org.apache.phoenix.util.DateUtil.parseTimestamp(DateUtil.java:175)
at org.apache.phoenix.schema.types.PTimestamp.toObject(PTimestamp.java:95)
at org.apache.phoenix.expression.LiteralExpression.newConstant(LiteralExpression.java:194)
at org.apache.phoenix.expression.LiteralExpression.newConstant(LiteralExpression.java:172)
at org.apache.phoenix.expression.LiteralExpression.newConstant(LiteralExpression.java:159)
at org.apache.phoenix.compile.UpsertCompiler$UpsertValuesCompiler.visit(UpsertCompiler.java:979)
at org.apache.phoenix.compile.UpsertCompiler$UpsertValuesCompiler.visit(UpsertCompiler.java:963)
at org.apache.phoenix.parse.BindParseNode.accept(BindParseNode.java:47)
at org.apache.phoenix.compile.UpsertCompiler.compile(UpsertCompiler.java:832)
at org.apache.phoenix.jdbc.PhoenixStatement$ExecutableUpsertStatement.compilePlan(PhoenixStatement.java:578)
at org.apache.phoenix.jdbc.PhoenixStatement$ExecutableUpsertStatement.compilePlan(PhoenixStatement.java:566)
at org.apache.phoenix.jdbc.PhoenixStatement$2.call(PhoenixStatement.java:331)
at org.apache.phoenix.jdbc.PhoenixStatement$2.call(PhoenixStatement.java:326)
at org.apache.phoenix.call.CallRunner.run(CallRunner.java:53)
at org.apache.phoenix.jdbc.PhoenixStatement.executeMutation(PhoenixStatement.java:324)
at org.apache.phoenix.jdbc.PhoenixStatement.execute(PhoenixStatement.java:245)
at org.apache.phoenix.jdbc.PhoenixPreparedStatement.execute(PhoenixPreparedStatement.java:172)
at org.apache.phoenix.jdbc.PhoenixPreparedStatement.execute(PhoenixPreparedStatement.java:177)
at org.apache.phoenix.mapreduce.PhoenixRecordWriter.write(PhoenixRecordWriter.java:79)
at org.apache.phoenix.mapreduce.PhoenixRecordWriter.write(PhoenixRecordWriter.java:39)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1$$anonfun$12$$anonfun$apply$4.apply$mcV$sp(PairRDDFunctions.scala:1113)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1$$anonfun$12$$anonfun$apply$4.apply(PairRDDFunctions.scala:1111)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1$$anonfun$12$$anonfun$apply$4.apply(PairRDDFunctions.scala:1111)
at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1251)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1$$anonfun$12.apply(PairRDDFunctions.scala:1119)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1$$anonfun$12.apply(PairRDDFunctions.scala:1091)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
at org.apache.spark.scheduler.Task.run(Task.scala:89)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.IllegalArgumentException: Invalid format: ""
at org.joda.time.format.DateTimeFormatter.parseDateTime(DateTimeFormatter.java:673)
at org.apache.phoenix.util.DateUtil$ISODateFormatParser.parseDateTime(DateUtil.java:295)
My code:
val myDF = sqlContext.createDataFrame(myRows, myStruct)
myDF.write
.format(sourcePhoenixSpark)
.mode("overwrite")
.options(Map("table" -> (myPhoenixNamespace + myTable), "zkUrl" -> myPhoenixZKUrl))
.save()
I am using phoenix-spark version 4.7.0-HBase-1.1. Any suggestion to solve the problem would be appreciated. Thanks
You are trying to process dirty data.
That error comes from here:
https://github.com/apache/phoenix/blob/master/phoenix-core/src/main/java/org/apache/phoenix/util/DateUtil.java#L301
Where it's trying to parse some string that is expected to be a Date in ISO format and the provided String is empty ("").
You need to prepare+clean your data before attempting to write it to storage.

Spark Streaming with large messages java.lang.OutOfMemoryError: Java heap space

I am using Spark Streaming 1.6.1 with Kafka0.9.0.1 (createStreams API) HDP 2.4.2, My use case sends large messages to Kafka Topics ranges from 5MB to 30 MB in such cases Spark Streaming fails to complete its job and crashes with below exception.I am doing a dataframe operation and saving on HDFS in csv format, below is my code snippet
Reading from Kafka Topic:
val lines = KafkaUtils.createStream[String, String, StringDecoder, StringDecoder](ssc, kafkaParams, topicMap, StorageLevel.MEMORY_AND_DISK_SER_2/*MEMORY_ONLY_SER_2*/).map(_._2)
Writing on HDFS:
val hdfsDF: DataFrame = getDF(sqlContext, eventDF, schema,topicName)
hdfsDF.show
hdfsDF.write
.format("com.databricks.spark.csv")
.option("header", "false")
.save(hdfsPath + "/" + "out_" + System.currentTimeMillis().toString())
16/11/11 12:12:35 WARN ReceiverTracker: Error reported by receiver for stream 0: Error handling message; exiting - java.lang.OutOfMemoryError: Java heap space
at java.lang.StringCoding$StringDecoder.decode(StringCoding.java:149)
at java.lang.StringCoding.decode(StringCoding.java:193)
at java.lang.String.<init>(String.java:426)
at java.lang.String.<init>(String.java:491)
at kafka.serializer.StringDecoder.fromBytes(Decoder.scala:50)
at kafka.serializer.StringDecoder.fromBytes(Decoder.scala:42)
at kafka.message.MessageAndMetadata.message(MessageAndMetadata.scala:32)
at org.apache.spark.streaming.kafka.KafkaReceiver$MessageHandler.run(KafkaInputDStream.scala:137)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
Followed by :
java.lang.Exception: Could not compute split, block input-0-1478610837000 not found
at org.apache.spark.rdd.BlockRDD.compute(BlockRDD.scala:51)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:313)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:277)
at org.apache.spark.rdd.MapPartitionsRDD.compute(MapPartitionsRDD.scala:38)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:313)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:277)
at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66)
at org.apache.spark.scheduler.Task.run(Task.scala:89)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
Driver stacktrace:

Implementing OLAP on titan/cassandra graph

I am using Titan 1.0.0 on top cassandra. I want to use OLAP services using SparkGraphComputer on the titan/cassandra graph. I have two questions
1) How to do it?
config:
https://github.com/thinkaurelius/titan/blob/titan10/titan-dist/src/assembly/static/conf/hadoop-graph/read-cassandra.properties
gremlin-code:
graph = GraphFactory.open('conf/hadoop-graph/read-cassandra.properties')
g = graph.traversal(computer(SparkGraphComputer))
g.V().count() //Here is the error
Error:
11:20:33 ERROR org.apache.spark.executor.Executor - Exception in task 3.0 in stage 0.0 (TID 3)
java.lang.RuntimeException: error communicating via Thrift
at org.apache.cassandra.hadoop.ColumnFamilyRecordReader$RowIterator.<init>(ColumnFamilyRecordReader.java:267)
at org.apache.cassandra.hadoop.ColumnFamilyRecordReader$RowIterator.<init>(ColumnFamilyRecordReader.java:215)
at org.apache.cassandra.hadoop.ColumnFamilyRecordReader$StaticRowIterator.<init>(ColumnFamilyRecordReader.java:331)
at org.apache.cassandra.hadoop.ColumnFamilyRecordReader$StaticRowIterator.<init>(ColumnFamilyRecordReader.java:331)
at org.apache.cassandra.hadoop.ColumnFamilyRecordReader.initialize(ColumnFamilyRecordReader.java:171)
at com.thinkaurelius.titan.hadoop.formats.cassandra.CassandraBinaryRecordReader.initialize(CassandraBinaryRecordReader.java:39)
at com.thinkaurelius.titan.hadoop.formats.util.GiraphRecordReader.initialize(GiraphRecordReader.java:38)
at org.apache.spark.rdd.NewHadoopRDD$$anon$1.<init>(NewHadoopRDD.scala:135)
at org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:107)
at org.apache.spark.rdd.NewHadoopRDD.compute(NewHadoopRDD.scala:69)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:280)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:247)
at org.apache.spark.rdd.MappedRDD.compute(MappedRDD.scala:31)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:280)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:247)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:68)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:41)
at org.apache.spark.scheduler.Task.run(Task.scala:56)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:200)
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)
Here is the full trace :
http://pastebin.com/CiuXjFB2
2) Why convert to hadoopGraph when the data is already stored on titan/cassandra?
references:
https://groups.google.com/forum/#!topic/gremlin-users/fVijONCxvSI

Is S3NativeFileSystem call killing my Pyspark Application on AWS EMR 4.6.0

My Spark application is failing when it has to access numerous CSV files (~1000 # 63MB each) from S3, and pipe them into a Spark RDD. The actual process of splitting up the CSV seems to work, but an extra function call to S3NativeFileSystem seems to be causing an error and the job to crash.
To begin, the following is my PySpark Application:
from pyspark import SparkContext
sc = SparkContext("local", "Simple App")
from pyspark.sql import SQLContext
sqlContext = SQLContext(sc)
import time
startTime = float(time.time())
dataPath = 's3://PATHTODIRECTORY/'
sc._jsc.hadoopConfiguration().set("fs.s3.awsAccessKeyId", "MYKEY")
sc._jsc.hadoopConfiguration().set("fs.s3.awsSecretAccessKey", "MYSECRETKEY")
def buildSchemaDF(tableName, columnList):
currentRDD = sc.textFile(dataPath + tableName).map(lambda line: line.split("|"))
currentDF = currentRDD.toDF(columnList)
return currentDF
loadStartTime = float(time.time())
lineitemDF = buildSchemaDF('lineitem*', ['l_orderkey','l_partkey','l_suppkey','l_linenumber','l_quantity','l_extendedprice','l_discount','l_tax','l_returnflag','l_linestatus','l_shipdate','l_commitdate','l_receiptdate','l_shipinstruct','l_shipmode','l_comment'])
lineitemDF.registerTempTable("lineitem")
loadTimeElapsed = float(time.time()) - loadStartTime
queryStartTime = float(time.time())
qstr = """
SELECT
lineitem.l_returnflag,
lineitem.l_linestatus,
sum(l_quantity) as sum_qty,
sum(l_extendedprice) as sum_base_price,
sum(l_discount) as sum_disc,
sum(l_tax) as sum_tax,
avg(l_quantity) as avg_qty,
avg(l_extendedprice) as avg_price,
avg(l_discount) as avg_disc,
count(l_orderkey) as count_order
FROM
lineitem
WHERE
l_shipdate <= '19981001'
GROUP BY
l_returnflag,
l_linestatus
ORDER BY
l_returnflag,
l_linestatus
"""
tpch1DF = sqlContext.sql(qstr)
queryTimeElapsed = float(time.time()) - queryStartTime
totalTimeElapsed = float(time.time()) - startTime
tpch1DF.show()
queryResults = [qstr, loadTimeElapsed, queryTimeElapsed, totalTimeElapsed]
distData = sc.parallelize(queryResults)
distData.saveAsTextFile(dataPath + 'queryResults.csv')
print 'Load Time: ' + str(loadTimeElapsed)
print 'Query Time: ' + str(queryTimeElapsed)
print 'Total Time: ' + str(totalTimeElapsed)
To take it step by step I start off by spinning up a Spark EMR Cluster with the following AWS CLI command (carriage returns added for readability):
aws emr create-cluster --name "Big TPCH Spark cluster2" --release-label emr-4.6.0
--applications Name=Spark --ec2-attributes KeyName=blazing-test-aws
--log-uri s3://aws-logs-132950491118-us-west-2/elasticmapreduce/j-1WZ39GFS3IX49/
--instance-type m3.2xlarge --instance-count 6 --use-default-roles
After the EMR cluster finishes provisioning I then copy over my Pyspark application onto the master node at '/home/hadoop/pysparkApp.py'. With it copied over I'm able to add the Step for spark-submit.
aws emr add-steps --cluster-id j-1DQJ8BDL1394N --steps
Type=spark,Name=SparkTPCHTests,Args=[--deploy-mode,cluster,-
conf,spark.yarn.submit.waitAppCompletion=true,--num-executors,5,--executor
cores,5,--executor memory,20g,/home/hadoop/tpchSpark.py]
,ActionOnFailure=CONTINUE
Now if I run this step over only a few of the aforementioned CSV files the final results will be generated, but the script will still claim to have failed.
I think it's associated with an extra call to S3NativeFileSystem, but I'm not certain. These are the Yarn log messages I'm getting which lead me to that conclusion. The first call appears to work just fine:
16/05/15 23:18:00 INFO HadoopRDD: Input split: s3://data-set-builder/splitLineItem2/lineitemad:0+64901757
16/05/15 23:18:00 INFO latency: StatusCode=[200], ServiceName=[Amazon S3], AWSRequestID=[ED8011CE4E1F6F18], ServiceEndpoint=[https://data-set-builder.s3-us-west-2.amazonaws.com], HttpClientPoolLeasedCount=0, RetryCapacityConsumed=0, RequestCount=1, HttpClientPoolPendingCount=0, HttpClientPoolAvailableCount=2, ClientExecuteTime=[77.956], HttpRequestTime=[77.183], HttpClientReceiveResponseTime=[20.028], RequestSigningTime=[0.229], CredentialsRequestTime=[0.003], ResponseProcessingTime=[0.128], HttpClientSendRequestTime=[0.35],
While the second one does not seem to execute properly, resulting in "Partial Results" (206 Error):
16/05/15 23:18:00 INFO S3NativeFileSystem: Opening 's3://data-set-builder/splitLineItem2/lineitemad' for reading
16/05/15 23:18:00 INFO latency: StatusCode=[206], ServiceName=[Amazon S3], AWSRequestID=[10BDDE61AE13AFBE], ServiceEndpoint=[https://data-set-builder.s3.amazonaws.com], HttpClientPoolLeasedCount=0, RetryCapacityConsumed=0, RequestCount=1, HttpClientPoolPendingCount=0, HttpClientPoolAvailableCount=2, Client Execute Time=[296.86], HttpRequestTime=[295.801], HttpClientReceiveResponseTime=[293.667], RequestSigningTime=[0.204], CredentialsRequestTime=[0.002], ResponseProcessingTime=[0.34], HttpClientSendRequestTime=[0.337],
16/05/15 23:18:02 INFO ApplicationMaster: Waiting for spark context initialization ...
I'm lost as to why it's even making the second call to S3NativeFileSystem when the first one appears to have responded effectively and even split the file. Is this something that is a product of my EMR configuration? I know S3Native has file limit issues and that a straight S3 call is optimal, which is what I've tried to do, but this call seems to be there no matter what I do. Please help!
Also, to add a few other error messages in my Yarn Log in case they are relevant.
1)
16/05/15 23:19:22 ERROR ApplicationMaster: SparkContext did not initialize after waiting for 100000 ms. Please check earlier log output for errors. Failing the application.
16/05/15 23:19:22 INFO ApplicationMaster: Final app status: FAILED, exitCode: 13, (reason: Timed out waiting for SparkContext.)
2)
16/05/15 23:19:22 ERROR DiskBlockObjectWriter: Uncaught exception while reverting partial writes to file /mnt/yarn/usercache/hadoop/appcache/application_1463354019776_0001/blockmgr-f847744b-c87a-442c-9135-57cae3d1f6f0/2b/temp_shuffle_3fe2e09e-f8e4-4e5d-ac96-1538bdc3b401
java.io.FileNotFoundException: /mnt/yarn/usercache/hadoop/appcache/application_1463354019776_0001/blockmgr-f847744b-c87a-442c-9135-57cae3d1f6f0/2b/temp_shuffle_3fe2e09e-f8e4-4e5d-ac96-1538bdc3b401 (No such file or directory)
at java.io.FileOutputStream.open(Native Method)
at java.io.FileOutputStream.<init>(FileOutputStream.java:221)
at org.apache.spark.storage.DiskBlockObjectWriter.revertPartialWritesAndClose(DiskBlockObjectWriter.scala:162)
at org.apache.spark.shuffle.sort.BypassMergeSortShuffleWriter.stop(BypassMergeSortShuffleWriter.java:226)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:79)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:41)
at org.apache.spark.scheduler.Task.run(Task.scala:89)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:745)
16/05/15 23:19:22 ERROR BypassMergeSortShuffleWriter: Error while deleting file /mnt/yarn/usercache/hadoop/appcache/application_1463354019776_0001/blockmgr-f847744b-c87a-442c-9135-57cae3d1f6f0/2b/temp_shuffle_3fe2e09e-f8e4-4e5d-ac96-1538bdc3b401
16/05/15 23:19:22 WARN TaskMemoryManager: leak 32.3 MB memory from org.apache.spark.unsafe.map.BytesToBytesMap#762be8fe
16/05/15 23:19:22 ERROR Executor: Managed memory leak detected; size = 33816576 bytes, TID = 14
16/05/15 23:19:22 ERROR Executor: Exception in task 13.0 in stage 1.0 (TID 14)
java.io.FileNotFoundException: /mnt/yarn/usercache/hadoop/appcache/application_1463354019776_0001/blockmgr-f847744b-c87a-442c-9135-57cae3d1f6f0/3a/temp_shuffle_b9001fca-bba9-400d-9bc4-c23c002e0aa9 (No such file or directory)
at java.io.FileOutputStream.open(Native Method)
at java.io.FileOutputStream.<init>(FileOutputStream.java:221)
at org.apache.spark.storage.DiskBlockObjectWriter.open(DiskBlockObjectWriter.scala:88)
at org.apache.spark.shuffle.sort.BypassMergeSortShuffleWriter.write(BypassMergeSortShuffleWriter.java:140)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:73)
at org.apache.spark.scheduler.ShuffleMapTask.runTask(ShuffleMapTask.scala:41)
at org.apache.spark.scheduler.Task.run(Task.scala:89)
at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:745)
Order of precedence for spark configurations is :
SparkContext (code/application) > Spark-submit > Spark-defaults.conf
So couple of things to point here -
Use YARN cluster as deploy mode and master in your spark submit command -
spark-submit --deploy-mode cluster --master yarn ...
OR
spark-submit --master yarn-cluster ...
Remove "local" string from line sc = SparkContext("local", "Simple App") in your code. Use conf = SparkConf().setAppName(appName)
sc = SparkContext(conf=conf) to initialize Spark context.
Ref - http://spark.apache.org/docs/latest/programming-guide.html

Resources