Spark is an in-memory open source cluster computing system allowing for fast iterative and interactive analytics. Spark utilizes Scala – a type-safe objected oriented language with functional properties that is fully interoperable with Java. For more information about Spark, please refer to http://spark-project.org. To test out Spark, you can install the stand-alone version on Mac OSX.
The first thing you will need to do is to install Scala 2.9.2 as Spark 0.6.1 is dependent on it. As of this posting, the current version of Scala is 2.10 but there are some issues with Spark 0.6.1 and Scala 2.10 as noted in this thread.
1) A handy way to installing Scala is to use Home Brew; please reference Installing Hadoop on OSX Lion (10.7) for more information on how to use Home Brew as well installing Hadoop on Mac OSX. It may be handy to install Hadoop so that way you can use Spark against HDFS as well.
2) The current Home Brew scala formula installs Scala 2.10 but you will need to use Scala 2.9.2. A quick way to do to this is to modify the scala.rb formula (/usr/local/Library/Formula/scala.rb) to install Scala 2.9.2.
require 'formula' class ScalaDocs < Formula homepage 'http://www.scala-lang.org/' url 'http://www.scala-lang.org/downloads/distrib/files/scala-docs-2.9.2.zip' sha1 '5bf44bd04b2b37976bde5d4a4c9bb6bcdeb10eb2' end class ScalaCompletion < Formula homepage 'http://www.scala-lang.org/' url 'https://raw.github.com/scala/scala-dist/27bc0c25145a83691e3678c7dda602e765e13413/completion.d/2.9.1/scala' version '2.9.1' sha1 'e2fd99fe31a9fb687a2deaf049265c605692c997' end class Scala < Formula homepage 'http://www.scala-lang.org/' url 'http://www.scala-lang.org/downloads/distrib/files/scala-2.9.2.tgz' #sha1 '87f605a186aa0e4435b302fb9af575513d29249a' sha1 '806fc1d91bda82d6a584172d7742531386ae68fb' option 'with-docs', 'Also install library documentation' def install rm_f Dir["bin/*.bat"] doc.install Dir['doc/*'] man1.install Dir['man/man1/*'] libexec.install Dir['*'] bin.install_symlink Dir["#{libexec}/bin/*"] ScalaCompletion.new.brew { (prefix/'etc/bash_completion.d').install 'scala' } ScalaDocs.new.brew { doc.install Dir['*'] } if build.include? 'with-docs' end end
3) Installing Scala via HomeBrew by typing the command in a bash terminal:
brew install scala
Upon running this command, scala will be located in /usr/local/Cellar/scala
Ensure you have Git for Mac installed (even if you have GitHub for Mac installed; need to install Git so you can run from the command line)
http://git-scm.com/download/mac
In my case, I have configured my .profile with the following:
# Java export JAVA_HOME=/Library/Java/JavaVirtualMachines/1.6.0_32-b05-417.jdk/Contents/Home # Scala export SCALA_HOME=/usr/local/Cellar/scala/2.9.2/libexec
1) Obtain the pre-built Spark 0.6.1 package at http://spark-project.org/downloads/. The direct link for the prebuilt package is:
http://github.com/downloads/mesos/spark/spark-0.6.1-prebuilt.tgz
2) Open up the tgz file and place it into a folder where you will install Spark. For example, I placed mine in the HomeBrew Cellar location, i.e.
/usr/local/Cellar/spark-0.6.1
Follow the instructions as per the README.MD in /usr/local/Cellar/spark-0.6.1
1) Run the Simple Build Tool (SBT) package from /usr/local/Cellar/spark-0.6.1
sbt/sbt package
2) Modify the conf/spark-env.sh
Ensure that SCALA_HOME variable has been set
export SCALA_HOME=/usr/local/Cellar/scala/2.9.2/libexec
From here, you can now run Spark examples. Just in case, run the conf/spark-env.sh to set the Scala enviornment variables.
conf/spark-env.sh ./run spark.examples.SparkLR local[2] ./run spark.examples.SparkPi local[4]
and to run the spark shell:
./spark-shell
where local indicates standalone (vs. EC2, cluster, mesos, etc.) and [x] is the number of cores.
Enjoy!
Pingback: Getting Hadoop and protobufs up and running with Elephant Bird on Mac OSX Mountain Lion | Denny Lee