Skip to content

SPARK-1492. Update Spark YARN docs to use spark-submit #601

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 3 commits into from
Closed

SPARK-1492. Update Spark YARN docs to use spark-submit #601

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
867a3ea
SPARK-1492. Update Spark YARN docs to use spark-submit
sryza Apr 30, 2014
be46d1f
Address feedback from Marcelo and Patrick
sryza May 1, 2014
5df1634
Address additional comments from Patrick.
sryza May 1, 2014
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
15 changes: 8 additions & 7 deletions docs/cluster-overview.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,7 +86,7 @@ the `--help` flag. Here are a few examples of common options:
--master local[8] \
my-app.jar

# Run on a Spark cluster
# Run on a Spark standalone cluster
./bin/spark-submit \
--class my.main.ClassName
--master spark://mycluster:7077 \
Expand Down Expand Up @@ -118,21 +118,22 @@ If you are ever unclear where configuration options are coming from. fine-graine
information can be printed by adding the `--verbose` option to `./spark-submit`.

### Advanced Dependency Management
When using `./bin/spark-submit` jars will be automatically transferred to the cluster. For many
users this is sufficient. However, advanced users can add jars by calling `addFile` or `addJar`
on an existing SparkContext. This can be used to distribute JAR files (Java/Scala) or .egg and
.zip libraries (Python) to executors. Spark uses the following URL scheme to allow different
When using `./bin/spark-submit` the app jar along with any jars included with the `--jars` option
will be automatically transferred to the cluster. `--jars` can also be used to distribute .egg and .zip
libraries for Python to executors. Spark uses the following URL scheme to allow different
strategies for disseminating jars:

- **file:** - Absolute paths and `file:/` URIs are served by the driver's HTTP file server, and
every executor pulls the file from the driver HTTP server
every executor pulls the file from the driver HTTP server.
- **hdfs:**, **http:**, **https:**, **ftp:** - these pull down files and JARs from the URI as expected
- **local:** - a URI starting with local:/ is expected to exist as a local file on each worker node. This
means that no network IO will be incurred, and works well for large files/JARs that are pushed to each worker,
or shared via NFS, GlusterFS, etc.

Note that JARs and files are copied to the working directory for each SparkContext on the executor nodes.
Over time this can use up a significant amount of space and will need to be cleaned up.
This can use up a significant amount of space over time and will need to be cleaned up. With YARN, cleanup
is handled automatically, and with Spark standalone, automatic cleanup can be configured with the
`spark.worker.cleanup.appDataTtl` property.

# Monitoring

Expand Down
115 changes: 28 additions & 87 deletions docs/running-on-yarn.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,27 +5,11 @@ title: Launching Spark on YARN

Support for running on [YARN (Hadoop
NextGen)](http://hadoop.apache.org/docs/r2.0.2-alpha/hadoop-yarn/hadoop-yarn-site/YARN.html)
was added to Spark in version 0.6.0, and improved in 0.7.0 and 0.8.0.

# Building a YARN-Enabled Assembly JAR

We need a consolidated Spark JAR (which bundles all the required dependencies) to run Spark jobs on a YARN cluster.
This can be built by setting the Hadoop version and `SPARK_YARN` environment variable, as follows:

SPARK_HADOOP_VERSION=2.0.5-alpha SPARK_YARN=true sbt/sbt assembly

The assembled JAR will be something like this:
`./assembly/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-assembly_{{site.SPARK_VERSION}}-hadoop2.0.5.jar`.

The build process now also supports new YARN versions (2.2.x). See below.
was added to Spark in version 0.6.0, and improved in subsequent releases.

# Preparations
Copy link
Contributor

@pwendell pwendell May 1, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It would also be nice to update this section if you have time. Right now it says you need to build the examples jar but again, almost all users will just download a pre-compiled version of spark that has examples in the lib/ directory.


- Building a YARN-enabled assembly (see above).
- The assembled jar can be installed into HDFS or used locally.
- Your application code must be packaged into a separate JAR file.

If you want to test out the YARN deployment mode, you can use the current Spark examples. A `spark-examples_{{site.SCALA_BINARY_VERSION}}-{{site.SPARK_VERSION}}` file can be generated by running `sbt/sbt assembly`. NOTE: since the documentation you're reading is for Spark version {{site.SPARK_VERSION}}, we are assuming here that you have downloaded Spark {{site.SPARK_VERSION}} or checked it out of source control. If you are using a different version of Spark, the version numbers in the jar generated by the sbt package command will obviously be different.
We need a consolidated Spark JAR (which bundles all the required dependencies) to run Spark jobs on a YARN cluster. The jar must be built with options that enable YARN support. To build this jar yourself, refer to the [building with maven guide](building-with-maven.html).

# Configuration

Expand All @@ -44,86 +28,47 @@ System Properties:
* `spark.yarn.max.executor.failures`, the maximum number of executor failures before failing the application. Default is the number of executors requested times 2 with minimum of 3.
* `spark.yarn.historyServer.address`, the address of the Spark history server (i.e. host.com:18080). The address should not contain a scheme (http://). Defaults to not being set since the history server is an optional service. This address is given to the Yarn ResourceManager when the Spark application finishes to link the application from the ResourceManager UI to the Spark history server UI.

By default, Spark on YARN will use a Spark jar installed locally, but the Spark jar can also be in a world-readable location on HDFS. This allows YARN to cache it on nodes so that it doesn't need to be distributed each time an application runs. To point to a jar on HDFS, export SPARK_JAR=hdfs:///some/path.

# Launching Spark on YARN

Ensure that HADOOP_CONF_DIR or YARN_CONF_DIR points to the directory which contains the (client side) configuration files for the Hadoop cluster.
These configs are used to connect to the cluster, write to the dfs, and connect to the YARN ResourceManager.
These configs are used to write to the dfs and connect to the YARN ResourceManager.

There are two deploy modes that can be used to launch Spark applications on YARN. In yarn-cluster mode, the Spark driver runs inside an application master process which is managed by YARN on the cluster, and the client can go away after initiating the application. In yarn-client mode, the driver runs in the client process, and the application master is only used for requesting resources from YARN.

Unlike in Spark standalone and Mesos mode, in which the master's address is specified in the "master" parameter, in YARN mode the ResourceManager's address is picked up from the Hadoop configuration. Thus, the master parameter is simply "yarn-client" or "yarn-cluster".

The spark-submit script described in the [cluster mode overview](cluster-overview.html) provides the most straightforward way to submit a compiled Spark application to YARN in either deploy mode. For info on the lower-level invocations it uses, read ahead. For running spark-shell against YARN, skip down to the yarn-client section.

## Launching a Spark application with yarn-cluster mode.

The command to launch the Spark application on the cluster is as follows:

SPARK_JAR=<SPARK_ASSEMBLY_JAR_FILE> ./bin/spark-class org.apache.spark.deploy.yarn.Client \
--jar <YOUR_APP_JAR_FILE> \
--class <APP_MAIN_CLASS> \
--arg <APP_MAIN_ARGUMENT> \
--num-executors <NUMBER_OF_EXECUTOR_PROCESSES> \
--driver-memory <MEMORY_FOR_ApplicationMaster> \
--executor-memory <MEMORY_PER_EXECUTOR> \
--executor-cores <CORES_PER_EXECUTOR> \
--name <application_name> \
--queue <queue_name> \
--addJars <any_local_files_used_in_SparkContext.addJar> \
--files <files_for_distributed_cache> \
--archives <archives_for_distributed_cache>

To pass multiple arguments the "arg" option can be specified multiple times. For example:

# Build the Spark assembly JAR and the Spark examples JAR
$ SPARK_HADOOP_VERSION=2.0.5-alpha SPARK_YARN=true sbt/sbt assembly

# Configure logging
$ cp conf/log4j.properties.template conf/log4j.properties

# Submit Spark's ApplicationMaster to YARN's ResourceManager, and instruct Spark to run the SparkPi example
$ SPARK_JAR=./assembly/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-assembly-{{site.SPARK_VERSION}}-hadoop2.0.5-alpha.jar \
./bin/spark-class org.apache.spark.deploy.yarn.Client \
--jar examples/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-examples-assembly-{{site.SPARK_VERSION}}.jar \
--class org.apache.spark.examples.SparkPi \
--arg yarn-cluster \
--arg 5 \
--num-executors 3 \
--driver-memory 4g \
--executor-memory 2g \
--executor-cores 1

The above starts a YARN client program which starts the default Application Master. Then SparkPi will be run as a child thread of Application Master. The client will periodically poll the Application Master for status updates and display them in the console. The client will exit once your application has finished running. Refer to the "Viewing Logs" section below for how to see driver and executor logs.

Because the application is run on a remote machine where the Application Master is running, applications that involve local interaction, such as spark-shell, will not work.

## Launching a Spark application with yarn-client mode.

With yarn-client mode, the application will be launched locally, just like running an application or spark-shell on Local / Mesos / Standalone client mode. The launch method is also the same, just make sure to specify the master URL as "yarn-client". You also need to export the env value for SPARK_JAR.
To launch a Spark application in yarn-cluster mode:

Configuration in yarn-client mode:
./bin/spark-submit --class path.to.your.Class --master yarn-cluster [options] <app jar> [app options]
Copy link
Contributor

@vanzin vanzin Apr 30, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This works, but I thought the preferred way was:

--master yarn --deploy-mode [client|cluster]


For example:

In order to tune executor cores/number/memory etc., you need to export environment variables or add them to the spark configuration file (./conf/spark_env.sh). The following are the list of options.
$ ./bin/spark-submit --class org.apache.spark.examples.SparkPi \
--master yarn-cluster \
--num-executors 3 \
--driver-memory 4g \
--executor-memory 2g \
--executor-cores 1
examples/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-examples-assembly-{{site.SPARK_VERSION}}.jar \
yarn-cluster 5

* `SPARK_EXECUTOR_INSTANCES`, Number of executors to start (Default: 2)
* `SPARK_EXECUTOR_CORES`, Number of cores per executor (Default: 1).
* `SPARK_EXECUTOR_MEMORY`, Memory per executor (e.g. 1000M, 2G) (Default: 1G)
* `SPARK_DRIVER_MEMORY`, Memory for driver (e.g. 1000M, 2G) (Default: 512 Mb)
* `SPARK_YARN_APP_NAME`, The name of your application (Default: Spark)
* `SPARK_YARN_QUEUE`, The YARN queue to use for allocation requests (Default: 'default')
* `SPARK_YARN_DIST_FILES`, Comma separated list of files to be distributed with the job.
* `SPARK_YARN_DIST_ARCHIVES`, Comma separated list of archives to be distributed with the job.
The above starts a YARN client program which starts the default Application Master. Then SparkPi will be run as a child thread of Application Master. The client will periodically poll the Application Master for status updates and display them in the console. The client will exit once your application has finished running. Refer to the "Viewing Logs" section below for how to see driver and executor logs.

For example:
To launch a Spark application in yarn-client mode, do the same, but replace "yarn-cluster" with "yarn-client". To run spark-shell:

SPARK_JAR=./assembly/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-assembly-{{site.SPARK_VERSION}}-hadoop2.0.5-alpha.jar \
./bin/run-example org.apache.spark.examples.SparkPi yarn-client
$ MASTER=yarn-client ./bin/spark-shell

or
## Adding additional jars

SPARK_JAR=./assembly/target/scala-{{site.SCALA_BINARY_VERSION}}/spark-assembly-{{site.SPARK_VERSION}}-hadoop2.0.5-alpha.jar \
MASTER=yarn-client ./bin/spark-shell
In yarn-cluster mode, the driver runs on a different machine than the client, so SparkContext.addJar won't work out of the box with files that are local to the client. To make files on the client available to SparkContext.addJar, include them with the `--jars` option in the launch command.

$ ./bin/spark-submit --class my.main.Class \
--master yarn-cluster \
--jars my-other-jar.jar,my-other-other-jar.jar
my-main-jar.jar
yarn-cluster 5

# Viewing logs

Expand All @@ -135,13 +80,9 @@ will print out the contents of all log files from all containers from the given

When log aggregation isn't turned on, logs are retained locally on each machine under YARN_APP_LOGS_DIR, which is usually configured to /tmp/logs or $HADOOP_HOME/logs/userlogs depending on the Hadoop version and installation. Viewing logs for a container requires going to the host that contains them and looking in this directory. Subdirectories organize log files by application ID and container ID.

# Building Spark for Hadoop/YARN 2.2.x

See [Building Spark with Maven](building-with-maven.html) for instructions on how to build Spark using Maven.

# Important notes

- Before Hadoop 2.2, YARN does not support cores in container resource requests. Thus, when running against an earlier version, the numbers of cores given via command line arguments cannot be passed to YARN. Whether core requests are honored in scheduling decisions depends on which scheduler is in use and how it is configured.
- The local directories used by Spark executors will be the local directories configured for YARN (Hadoop YARN config yarn.nodemanager.local-dirs). If the user specifies spark.local.dir, it will be ignored.
- The --files and --archives options support specifying file names with the # similar to Hadoop. For example you can specify: --files localtest.txt#appSees.txt and this will upload the file you have locally named localtest.txt into HDFS but this will be linked to by the name appSees.txt, and your application should use the name as appSees.txt to reference it when running on YARN.
- The --addJars option allows the SparkContext.addJar function to work if you are using it with local files. It does not need to be used if you are using it with HDFS, HTTP, HTTPS, or FTP files.
- The --jars option allows the SparkContext.addJar function to work if you are using it with local files and running in yarn-cluster mode. It does not need to be used if you are using it with HDFS, HTTP, HTTPS, or FTP files.