Apache Spark - A unified analytics engine for large-scale data processing

Go to file

Matei Zaharia 628bb5ca7f Allow null keys in Spark's reduce and group by		2012-07-12 18:36:02 -07:00
bagel/src	Merge branch 'master' into dev	2012-06-15 23:47:11 -07:00
conf	Further fixes to how Mesos is found and used	2012-03-17 13:39:14 -07:00
core	Allow null keys in Spark's reduce and group by	2012-07-12 18:36:02 -07:00
ec2	Increase the default wait time for EC2 clusters to 2 minutes.	2012-06-29 16:25:40 -07:00
examples/src/main/scala/spark/examples	Add System.exit(0) at the end of all the example programs.	2012-06-05 23:31:28 -07:00
project	Initial framework to get a master and web UI up.	2012-06-30 14:45:55 -07:00
repl	Some refactoring to make cluster scheduler pluggable.	2012-07-06 15:23:26 -07:00
sbt	Merge branch 'dev' of github.com:mesos/spark into dev	2012-06-15 23:36:30 -07:00
.gitignore	Ignore some IDE files	2012-06-28 23:52:33 -07:00
kmeans_data.txt	Fixed bugs	2012-01-09 11:59:52 -08:00
LICENSE	Added BSD license	2010-12-07 10:32:17 -08:00
lr_data.txt	Test commit	2012-02-06 09:58:06 -08:00
README.md	Documentation	2012-03-17 13:51:29 -07:00
run	More work to allow Spark to run on the standalone deploy cluster.	2012-07-08 14:00:04 -07:00
spark-executor	Further refactoring, and start of a standalone scheduler backend	2012-07-06 17:56:44 -07:00
spark-shell	More work to allow Spark to run on the standalone deploy cluster.	2012-07-08 14:00:04 -07:00

README.md

Spark

Lightning-Fast Cluster Computing - http://www.spark-project.org/

Online Documentation

You can find the latest Spark documentation, including a programming guide, on the project wiki at http://github.com/mesos/spark/wiki. This file only contains basic setup instructions.

Building

Spark requires Scala 2.9.1. This version has been tested with 2.9.1.final.

The project is built using Simple Build Tool (SBT), which is packaged with it. To build Spark and its example programs, run:

sbt/sbt compile

To run Spark, you will need to have Scala's bin in your PATH, or you will need to set the SCALA_HOME environment variable to point to where you've installed Scala. Scala must be accessible through one of these methods on Mesos slave nodes as well as on the master.

To run one of the examples, use ./run <class> <params>. For example:

./run spark.examples.SparkLR local[2]

will run the Logistic Regression example locally on 2 CPUs.

Each of the example programs prints usage help if no params are given.

All of the Spark samples take a <host> parameter that is the Mesos master to connect to. This can be a Mesos URL, or "local" to run locally with one thread, or "local[N]" to run locally with N threads.

A Note About Hadoop

Spark uses the Hadoop core library to talk to HDFS and other Hadoop-supported storage systems. Because the HDFS API has changed in different versions of Hadoop, you must build Spark against the same version that your cluster runs. You can change the version by setting the HADOOP_VERSION variable at the top of project/SparkBuild.scala, then rebuilding Spark.

Configuration

Spark can be configured through two files: conf/java-opts and conf/spark-env.sh.

In java-opts, you can add flags to be passed to the JVM when running Spark.

In spark-env.sh, you can set any environment variables you wish to be available when running Spark programs, such as PATH, SCALA_HOME, etc. There are also several Spark-specific variables you can set:

SPARK_CLASSPATH: Extra entries to be added to the classpath, separated by ":".
SPARK_MEM: Memory for Spark to use, in the format used by java's -Xmx option (for example, -Xmx200m means 200 MB, -Xmx1g means 1 GB, etc).
SPARK_LIBRARY_PATH: Extra entries to add to java.library.path for locating shared libraries.
SPARK_JAVA_OPTS: Extra options to pass to JVM.
MESOS_NATIVE_LIBRARY: Your Mesos library, if you want to run on a Mesos cluster. For example, this might be /usr/local/lib/libmesos.so on Linux.

Note that spark-env.sh must be a shell script (it must be executable and start with a #! header to specify the shell to use).