we were using spark 1.3.1 and launching our spark jobs on yarn-client mode programmatically via creating a sparkConf and sparkContext object manually. It was inspired from spark self-contained application example here:
https://spark.apache.org/docs/1.5.2/quick-start.html#self-contained-applications\
Only additional configuration we would provide would be all related to yarn like executor instance, cores etc.
However after upgrading to spark 1.5.2 above application breaks on a line val sparkContext = new SparkContext(sparkConf)
It throws following in driver application:
16/01/28 17:38:35 ERROR util.Utils: Uncaught exception in thread main
java.lang.NullPointerException
at org.apache.spark.network.netty.NettyBlockTransferService.close(NettyBlockTransferService.scala:152)
at org.apache.spark.storage.BlockManager.stop(BlockManager.scala:1228)
at org.apache.spark.SparkEnv.stop(SparkEnv.scala:100)
at org.apache.spark.SparkContext$$anonfun$stop$12.apply$mcV$sp(SparkContext.scala:1749)
at org.apache.spark.util.Utils$.tryLogNonFatalError(Utils.scala:1185)
at org.apache.spark.SparkContext.stop(SparkContext.scala:1748)
at org.apache.spark.SparkContext.<init>(SparkContext.scala:593)
So is this approach still supposed to work? Or do I must use SparkLauncher class with spark 1.5.2 to launch spark job programmatically on yarn-client mode?
Related
I want to share an interesting error I've caught up recently:
Exception in thread "main" java.lang.NoSuchFieldError: TOKEN_KIND
at org.apache.hadoop.crypto.key.kms.KMSClientProvider$KMSTokenRenewer.handleKind(KMSClientProvider.java:166)
at org.apache.hadoop.security.token.Token.getRenewer(Token.java:351)
at org.apache.hadoop.security.token.Token.renew(Token.java:377)
at org.apache.spark.deploy.security.HadoopFSCredentialProvider$$anonfun$getTokenRenewalInterval$1$$anonfun$5$$anonfun$apply$1.apply$mcJ$sp(HadoopFSDelegationTokeProvider.scala:119)
I was trying to spark2-submit a job to a remote driver host on Cloudera cluster like this:
spark = SparkSession.builder
.master("yarn")
.config("cluster")
.config("spark.driver.host", "remote_driver_host")
.config("spark.yarn.keytab", "path_to_pricnipar.keytab")
.config("spark.yarn.principal", "principal.name") \
.config("spark.driver.bindAddress", "0.0.0.0") \
.getOrCreate()
The Apache spark and Hadoop versions on Cloudera cluster are: 2.3.0 and 2.6.0 accordingly.
So the cause of issue was quite trivial, it is spark local binaries vs remote spark driver version mismatch.
Locally I had installed spark 2.4.5 and on Cloudera it was 2.3.0, after aligning the versions to 2.3.0, the issue resolved and the spark job completed successfully.
Scala version: 2.11.12
Spark version: 2.4.0
emr-5.23.0
Get the following when running the below command to create an Amazon EMR cluster
spark-submit --class etl.SparkDataProcessor --master yarn --deploy-mode cluster --conf spark.yarn.appMasterEnv.ETL_NAME=foo --conf spark.yarn.appMasterEnv.ETL_SPARK_MASTER=yarn --conf spark.yarn.appMasterEnv.ETL_AWS_ACCESS_KEY_ID=123 --conf spark.yarn.appMasterEnv.ETL_AWS_SECRET_ACCESS_KEY=abc MY-Tool.jar
Exception
ERROR ApplicationMaster: Uncaught exception:
java.lang.IllegalStateException: User did not initialize spark context!
at org.apache.spark.deploy.yarn.ApplicationMaster.runDriver(ApplicationMaster.scala:485)
at org.apache.spark.deploy.yarn.ApplicationMaster.org$apache$spark$deploy$yarn$ApplicationMaster$$runImpl(ApplicationMaster.scala:305)
at org.apache.spark.deploy.yarn.ApplicationMaster$$anonfun$run$1.apply$mcV$sp(ApplicationMaster.scala:245)
at org.apache.spark.deploy.yarn.ApplicationMaster$$anonfun$run$1.apply(ApplicationMaster.scala:245)
at org.apache.spark.deploy.yarn.ApplicationMaster$$anonfun$run$1.apply(ApplicationMaster.scala:245)
at org.apache.spark.deploy.yarn.ApplicationMaster$$anon$3.run(ApplicationMaster.scala:773)
at java.security.AccessController.doPrivileged(Native Method)
at javax.security.auth.Subject.doAs(Subject.java:422)
at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1844)
at org.apache.spark.deploy.yarn.ApplicationMaster.doAsUser(ApplicationMaster.scala:772)
at org.apache.spark.deploy.yarn.ApplicationMaster.run(ApplicationMaster.scala:244)
at org.apache.spark.deploy.yarn.ApplicationMaster$.main(ApplicationMaster.scala:797)
at org.apache.spark.deploy.yarn.ApplicationMaster.main(ApplicationMaster.scala)
How I create my spark session (where sparkMaster = yarn)
lazy val spark: SparkSession = {
val logger: Logger = Logger.getLogger("etl");
val sparkAppName = EnvConfig.ETL_NAME
val sparkMaster = EnvConfig.ETL_SPARK_MASTER
val sparkInstance = SparkSession
.builder()
.appName(sparkAppName)
.master(sparkMaster)
.getOrCreate()
val hadoopConf = sparkInstance.sparkContext.hadoopConfiguration
hadoopConf.set("fs.s3.impl", "org.apache.hadoop.fs.s3a.S3AFileSystem")
hadoopConf.set("fs.s3a.access.key", EnvConfig.ETL_AWS_ACCESS_KEY_ID)
hadoopConf.set("fs.s3a.secret.key", EnvConfig.ETL_AWS_SECRET_ACCESS_KEY)
logger.info("Created My SparkSession")
logger.info(s"Spark Application Name: $sparkAppName")
logger.info(s"Spark Master: $sparkMaster")
sparkInstance
}
UPDATE:
I determined that due to the application logic, in certain cases, we did not initialize the spark session. Because of this, it seems that when the cluster terminates, it also tries to do something with the session (perhaps close it) and is thus failing. Now that I have figured out this issue, the application runs but never actually completes. Currently, it seems to be hanging in a particular part involving spark when running in cluster mode:
val data: DataFrame = spark.read
.option("header", "true")
.option("inferSchema", "true")
.csv(s"s3://$csvPath/$fileKey")
.toDF()
20/03/16 18:38:35 INFO Client: Application report for application_1584324418613_0031 (state: RUNNING)
AFAIK EnvConfig.ETL_AWS_ACCESS_KEY_ID and ETL_AWS_SECRET_ACCESS_KEY are not getting populated due to which sparksession cant be instanciated with null or empty values . try to print and debug the values.
also reading the properties from --conf spark.xxx
should be like this example. I hope you are following this...
spark.sparkContext.getConf.getOption("spark. ETL_AWS_ACCESS_KEY_ID")
once you check that, this example way should work...
/**
* Hadoop-AWS Configuration
*/
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3a.proxy.host", proxyHost)
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3a.proxy.port", proxyPort)
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3a.aws.credentials.provider", "com.amazonaws.auth.DefaultAWSCredentialsProviderChain")
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3.impl", "org.apache.hadoop.fs.s3native.NativeS3FileSystem")
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3a.impl", "org.apache.hadoop.fs.s3a.S3AFileSystem")
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3a.server-side-encryption-algorithm", "AES256")
sparkSession.sparkContext.hadoopConfiguration.set("fs.s3n.server-side-encryption-algorithm", "AES256")
sparkSession.sparkContext.hadoopConfiguration.set("spark.hadoop.fs.s3a.impl", "org.apache.hadoop.fs.s3a.S3AFileSystem
another thing is, use
--master yarn or --master local[*] you can use instead of
-conf spark.yarn.appMasterEnv.ETL_SPARK_MASTER=yarn
UPDATE :
--conf spark.driver.port=20002 may solve this issue. where 20002 is orbitary port.. seems like its waiting for the particular port for some time and its retrying for some time and its failing with the exception you got.
I got this idea by walking through the Sparks application master code from here
and comment This a bit hacky, but we need to wait until the spark.driver.port property has been set by the Thread executing the user class.
you can try this and let me know.
Further reading : Apache Spark : How to change the port the Spark driver listens to
In my case (after resolving the application issues), I needed to include core AND task node types when deploying in cluster mode.
a newbie in apache spark here! I am using Spark 2.4.0 and Scala version 2.11.12, and I'm trying to run the following code in my spark shell -
import org.apache.spark.sql.SparkSession
import spark.implicits._
var df = spark.read.json("storesales.json")
df.createOrReplaceTempView("storesales")
spark.sql("SELECT * FROM storesales")
And I get the following error -
2018-12-18 07:05:03 WARN Hive:168 - Failed to access metastore. This class should not accessed in runtime.
org.apache.hadoop.hive.ql.metadata.HiveException: java.lang.RuntimeException: Unable to instantiate org.apache.hadoop.
hive.ql.metadata.SessionHiveMetaStoreClient
at org.apache.hadoop.hive.ql.metadata.Hive.getAllDatabases(Hive.java:1236)
at org.apache.hadoop.hive.ql.metadata.Hive.reloadFunctions(Hive.java:174)
at org.apache.hadoop.hive.ql.metadata.Hive.<clinit>(Hive.java:166)
at org.apache.hadoop.hive.ql.session.SessionState.start(SessionState.java:503)
at org.apache.spark.sql.hive.client.HiveClientImpl.newState(HiveClientImpl.scala:183)
at org.apache.spark.sql.hive.client.HiveClientImpl.<init>(HiveClientImpl.scala:117)
at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
at java.base/jdk.internal.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java
:62)
I also saw this Issues trying out example in Spark-shell and as per the accepted answer, I have tried to start my spark shell like so,
~/spark-2.4.0-bin-hadoop2.7/bin/spark-shell --conf spark.sql.warehouse.dir=file:///tmp/spark-warehouse, however, it did not help and the issue persists.
I am creating a self-contained Scala program that uses Spark for parallelization in some parts. In my specific situation, the Spark cluster is available through mesos.
I create spark context like this:
val conf = new SparkConf().setMaster("mesos://zk://<mesos-url1>,<mesos-url2>/spark/mesos-rtspark").setAppName("foo")
val sc = new SparkContext(conf)
I found out from searching around that you have to specify MESOS_NATIVE_JAVA_LIBRARY env var to point to the libmesos library, so when running my Scala program I do this:
MESOS_NATIVE_JAVA_LIBRARY=/usr/local/lib/libmesos.dylib sbt run
But, this results in a SparkException:
ERROR SparkContext: Error initializing SparkContext.
org.apache.spark.SparkException: Could not parse Master URL: 'mesos://zk://<mesos-url1>,<mesos-url2>/spark/mesos-rtspark'
At the same time, using spark-submit seems to work fine after exporting the MESOS_NATIVE_JAVA_LIBRARY env var.
MESOS_NATIVE_JAVA_LIBRARY=/usr/local/lib/libmesos.dylib spark-submit --class <MAIN CLASS> ./target/scala-2.10/<APP_JAR>.jar
Why?
How can I make the standalone program run like spark-submit?
Add spark-mesos jar to your classpath.
I used this code
My error is:
Using Spark's default log4j profile: org/apache/spark/log4j-defaults.properties
17/02/03 20:39:24 INFO SparkContext: Running Spark version 2.1.0
17/02/03 20:39:25 WARN NativeCodeLoader: Unable to load native-hadoop
library for your platform... using builtin-java classes where applicable
17/02/03 20:39:25 WARN SparkConf: Detected deprecated memory fraction
settings: [spark.storage.memoryFraction]. As of Spark 1.6, execution and
storage memory management are unified. All memory fractions used in the old
model are now deprecated and no longer read. If you wish to use the old
memory management, you may explicitly enable `spark.memory.useLegacyMode`
(not recommended).
17/02/03 20:39:25 ERROR SparkContext: Error initializing SparkContext.
org.apache.spark.SparkException: A master URL must be set in your
configuration
at org.apache.spark.SparkContext.<init>(SparkContext.scala:379)
at PCA$.main(PCA.scala:26)
at PCA.main(PCA.scala)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at com.intellij.rt.execution.application.AppMain.main(AppMain.java:144)
17/02/03 20:39:25 INFO SparkContext: Successfully stopped SparkContext
Exception in thread "main" org.apache.spark.SparkException: A master URL must be set in your configuration
at org.apache.spark.SparkContext.<init>(SparkContext.scala:379)
at PCA$.main(PCA.scala:26)
at PCA.main(PCA.scala)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at com.intellij.rt.execution.application.AppMain.main(AppMain.java:144)
Process finished with exit code 1
If you are running spark stand alone then
val conf = new SparkConf().setMaster("spark://master") //missing
and you can pass parameter while submit job
spark-submit --master spark://master
If you are running spark local then
val conf = new SparkConf().setMaster("local[2]") //missing
you can pass parameter while submit job
spark-submit --master local
if you are running spark on yarn then
spark-submit --master yarn
Error message is pretty clear, you have to provide the address of the Spark Master node, either via the SparkContext or via spark-submit:
val conf =
new SparkConf()
.setAppName("ClusterScore")
.setMaster("spark://172.1.1.1:7077") // <--- This is what's missing
.set("spark.storage.memoryFraction", "1")
val sc = new SparkContext(conf)
SparkConf configuration = new SparkConf()
.setAppName("Your Application Name")
.setMaster("local");
val sc = new SparkContext(conf);
It will work...
Most probably you are using Spark 2.x API in Java.
Use code snippet like this to avoid this error. This is true when you are running Spark standalone on your computer using Shade plug-in which will import all the runtime libraries on your computer.
SparkSession spark = SparkSession.builder()
.appName("Spark-Demo")//assign a name to the spark application
.master("local[*]") //utilize all the available cores on local
.getOrCreate();