I have installed Hadoop2.7.1 with spark 1.4.1 on windows 8.1
When I execute below command
cd spark
spark-submit --class org.apache.spark.examples.SparkPi --master yarn --deploy-mode client lib/spark-examples*.jar 10
I get below error in JobHistoryServer log
Error: Could not find or load main class '-Dspark.externalBlockStore.folderName=spark-262c4697-ef0c-4042-af0c-8106b08574fb'
I did further debugging(along searching net) and could get hold of container cmd script where below sections(other lines are omitted) are given
...
#set CLASSPATH=C:/tmp/hadoop-xyz/nm-local-dir/usercache/xyz/appcache/application_1487502025818_0003/container_1487502025818_0003_02_000001/classpath-3207656532274684591.jar
...
#call %JAVA_HOME%/bin/java -server -Xmx512m -Djava.io.tmpdir=%PWD%/tmp '-Dspark.fileserver.uri=http://192.168.1.2:34814' '-Dspark.app.name=Spark shell' '-Dspark.driver.port=34810' '-Dspark.repl.class.uri=http://192.168.1.2:34785' '-Dspark.driver.host=192.168.1.2' '-Dspark.externalBlockStore.folderName=spark-dd9f3f84-6cf4-4ff8-b0f6-7ff84daf74bc' '-Dspark.master=yarn-client' '-Dspark.driver.appUIAddress=http://192.168.1.2:4040' '-Dspark.jars=' '-Dspark.executor.id=driver' -Dspark.yarn.app.container.log.dir=/dep/logs/userlogs/application_1487502025818_0003/container_1487502025818_0003_02_000001 org.apache.spark.deploy.yarn.ExecutorLauncher --arg '192.168.1.2:34810' --executor-memory 1024m --executor-cores 1 --num-executors 2 1> /dep/logs/userlogs/application_1487502025818_0003/container_1487502025818_0003_02_000001/stdout 2> /dep/logs/userlogs/application_1487502025818_0003/container_1487502025818_0003_02_000001/stderr
I check relevant files for CLASSPATH, looks OK. The main class org.apache.spark.deploy.yarn.ExecutorLauncher is available in spark assembly jar which is part of container jar
So, what could be the issue here? I searched net and found many discussions, but are for unix variants, but not many for Windows. I am wondering whether spark submit really works on Windows (yarn-client mode only, standalone cluster mode works) without any special setup!!!
BTW, if I run the above java command from cmd.exe command prompt, I get the same error as all command line arguments are quoted with single quote instead of double quotes(making these double quotes work!!), so is this a bug
Note spark-shell also fails (in yarn mode) and but yarn jar ... command works
Looks like it was a defect in earlier version. With latest Hadoop 2.7.3 with spark 2.1.0, it is working correctly.!!! Could not find any reference though.
Related
I am looking for a guide regarding how to install spark on an existing virtual yarn cluster.
I have a yarn cluster consisting of two nodes, ran map-reduce job which worked perfect. Looked for results in log and everything is working fine.
Now I need to add the spark installation commands and configurations files in my vagrantfile. I can't find a good guide, could someone give me a good link ?
I used this guide for the yarn cluster
http://www.alexjf.net/blog/distributed-systems/hadoop-yarn-installation-definitive-guide/#single-node-installation
Thanks in advance!
I don't know about vagrant, but I have installed Spark on top of hadoop 2.6 (in the guide referred to as post-YARN) and I hope this helps.
Installing Spark on an existing hadoop is really easy, you just need to install it only on one machine. For that you have to download the one pre-built for your hadoop version from it's official website (I guess you can use the without hadoop version but you need to point it to the direction of hadoop binaries in your system). Then decompress it:
tar -xvf spark-2.0.0-bin-hadoop2.x.tgz -C /opt
Now you only need to set some environment variables. First in your ~/.bashrc (or ~/.zshrc) you can set SPARK_HOME and add it to your PATH if you want:
export SPARK_HOME=/opt/spark-2.0.0-bin-hadoop-2.x
export PATH=$PATH:$SPARK_HOME/bin
Also for this changes to take effect you can run:
source ~/.bashrc
Second you need to point Spark to your Hadoop configuartion directories. To do this set these two environmental variables in $SPARK_HOME/conf/spark-env.sh:
export HADOOP_CONF_DIR=[your-hadoop-conf-dir usually $HADOOP_PREFIX/etc/hadoop]
export YARN_CONF_DIR=[your-yarn-conf-dir usually the same as the last variable]
If this file doesn't exist, you can copy the contents of $SPARK_HOME/conf/spark-env.sh.template and start from there.
Now to start the shell in yarn mode you can run:
spark-shell --master yarn --deploy-mode client
(You can't run the shell in cluster deploy-mode)
----------- Update
I forgot to mention that you can also submit cluster jobs with this configuration like this (thanks #JulianCienfuegos):
spark-submit --master yarn --deploy-mode cluster project-spark.py
This way you can't see the output in the terminal, and the command exits as soon as the job is submitted (not completed).
You can also use --deploy-mode client to see the output right there in your terminal but just do this for testing, since the job gets canceled if the command is interrupted (e.g. you press Ctrl+C, or your session ends)
I met the following issue that Spark can not launch on Windows Yarn.
15/06/05 06:31:34 ERROR spark.SparkContext: Error initializing SparkContext.org.apache.spark.SparkException: Yarn application has already ended! It might have been killed or unable to launch application master.
at org.apache.spark.scheduler.cluster.YarnClientSchedulerBackend.waitForApplication(YarnClientSchedulerBackend.scala:114)
at org.apache.spark.scheduler.cluster.YarnClientSchedulerBackend.start(YarnClientSchedulerBackend.scala:59)
And when I drill down to Yarn, and get the error:
Error: Could not find or load main class '-Dspark.driver.memory=2G'
After investigating this issue, the root cause is:
In the YARN command generation part, single quotes (‘) will be
added surrounding some of the java options.
But Windows does not like the quotes for these options.
Similar to this issue https://issues.apache.org/jira/browse/SPARK-5754 .
Anyone know how to escape the logic working on Windows Yarn cluster?
Spark Submit Command:
%SPARK_HOME%\bin\spark-submit.cmd --jars ... ^
--class ....^
--master yarn-client ^
--driver-memory 10G ^
--executor-memory 20G ^
--executor-cores 6 ^
--num-executors 10 ^
QuasarNRT.jar 10 6
-Tao
I'm trying to run a spark application using bin/spark-submit. When I reference my application jar inside my local filesystem, it works. However, when I copied my application jar to a directory in hdfs, i get the following exception:
Warning: Skip remote jar hdfs://localhost:9000/user/hdfs/jars/simple-project-1.0-SNAPSHOT.jar.
java.lang.ClassNotFoundException: com.example.SimpleApp
Here's the command:
$ ./bin/spark-submit --class com.example.SimpleApp --master local hdfs://localhost:9000/user/hdfs/jars/simple-project-1.0-SNAPSHOT.jar
I'm using hadoop version 2.6.0, spark version 1.2.1
The only way it worked for me, when I was using
--master yarn-cluster
To make HDFS library accessible to spark-job , you have to run job in cluster mode.
$SPARK_HOME/bin/spark-submit \
--deploy-mode cluster \
--class <main_class> \
--master yarn-cluster \
hdfs://myhost:8020/user/root/myjar.jar
Also, There is Spark JIRA raised for client mode which is not supported yet.
SPARK-10643 :Support HDFS application download in client mode spark submit
There is a workaround. You could mount the directory in HDFS (which contains your application jar) as local directory.
I did the same (with azure blob storage, but it should be similar for HDFS)
example command for azure wasb
sudo mount -t cifs //{storageAccountName}.file.core.windows.net/{directoryName} {local directory path} -o vers=3.0,username={storageAccountName},password={storageAccountKey},dir_mode=0777,file_mode=0777
Now, in your spark submit command, you provide the path from the command above
$ ./bin/spark-submit --class com.example.SimpleApp --master local {local directory path}/simple-project-1.0-SNAPSHOT.jar
spark-submit --master spark://kssr-virtual-machine:7077 --deploy-mode client --executor-memory 1g hdfs://localhost:9000/user/wordcount.py
For me its working I am using Hadoop 3.3.1 & Spark 3.2.1. I am able to read the file from HDFS.
Yes, it has to be a local file. I think that's simply the answer.
I am trying to run a fat jar on a Spark cluster using Spark submit.
I made the cluster using "spark-ec2" executable in Spark bundle on AWS.
The command I am using to run the jar file is
bin/spark-submit --class edu.gatech.cse8803.main.Main --master yarn-cluster ../src1/big-data-hw2-assembly-1.0.jar
In the beginning it was giving me the error that at least one of the HADOOP_CONF_DIR or YARN_CONF_DIR environment variable must be set.
I didn't know what to set them to, so I used the following command
export HADOOP_CONF_DIR=/mapreduce/conf
Now the error has changed to
Could not load YARN classes. This copy of Spark may not have been compiled with YARN support.
Run with --help for usage help or --verbose for debug output
The home directory structure is as follows
ephemeral-hdfs hadoop-native mapreduce persistent-hdfs scala spark spark-ec2 src1 tachyon
I even set the YARN_CONF_DIR variable to the same value as HADOOP_CONF_DIR, but the error message is not changing. I am unable to find any documentation that highlights this issue, most of them just mention these two variables and give no further details.
You need to compile spark against Yarn to use it.
Follow the steps explained here: https://spark.apache.org/docs/latest/building-spark.html
Maven:
build/mvn -Pyarn -Phadoop-2.x -Dhadoop.version=2.x.x -DskipTests clean package
SBT:
build/sbt -Pyarn -Phadoop-2.x assembly
You can also download a pre-compiled version here: http://spark.apache.org/downloads.html (choose a "pre-built for Hadoop")
Download prebuilt spark which supports hadoop 2.X versions from https://spark.apache.org/downloads.html
The --master argument should be: --master spark://hostname:7077 where hostname is the name of your Spark master server. You can also specify this value as spark.master in the spark-defaults.conf file and leave out the --master argument when using Spark submit from the command line. Including the --master argument will override the value set (if exists) in the spark-defaults.conf file.
Reference: http://spark.apache.org/docs/1.3.0/configuration.html
I have been trying to get a Spark Streaming job, running on a EC2 instance to report to VisualVM using JMX.
As of now I have the following config file:
spark/conf/metrics.properties:
*.sink.jmx.class=org.apache.spark.metrics.sink.JmxSink
master.source.jvm.class=org.apache.spark.metrics.source.JvmSource
worker.source.jvm.class=org.apache.spark.metrics.source.JvmSource
driver.source.jvm.class=org.apache.spark.metrics.source.JvmSource
executor.source.jvm.class=org.apache.spark.metrics.source.JvmSource
And I start the spark streaming job like this:
(the -D bits I have added afterwards in the hopes of getting remote access to the ec2's jmx)
terminal:
spark/bin/spark-submit --class my.class.StarterApp --master local --deploy-mode client \
project-1.0-SNAPSHOT.jar \
-Dcom.sun.management.jmxremote \
-Dcom.sun.management.jmxremote.port=54321 \
-Dcom.sun.management.jmxremote.authenticate=false \
-Dcom.sun.management.jmxremote.ssl=false
There are two issues with the spark-submit command line:
local - you must not run Spark Standalone with local master URL because there will be no threads to run your computations (jobs) and you've got two, i.e. one for a receiver and another for the driver. You should see the following WARN in the logs:
WARN StreamingContext: spark.master should be set as local[n], n > 1
in local mode if you have receivers to get data, otherwise Spark jobs
will not get resources to process the received data.
-D options are not picked up by the JVM as they're given after the Spark Streaming application and effectively became its command-line arguments. Put them before project-1.0-SNAPSHOT.jar and start over (you have to fix the above issue first!)
spark-submit --conf "spark.driver.extraJavaOptions=-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=8090 -Dcom.sun.management.jmxremote.authenticate=false -Dcom.sun.management.jmxremote.ssl=false"/path/example/src/main/python/pi.py 10000
Notes:the configurations format : --conf "params" . tested under spark 2.+