YARN - SparkPi was updated to not take in master as an argument; we should update the docs to reflect that. - The default YARN build guide should be in maven, not sbt. - This PR also adds a paragraph on steps to debug a YARN application. Standalone - Emphasize spark-submit more. Right now it's one small paragraph preceding the legacy way of launching through `org.apache.spark.deploy.Client`. - The way we set configurations / environment variables according to the old docs is outdated. This needs to reflect changes introduced by the Spark configuration changes we made. In general, this PR also adds a little more documentation on the new spark-shell, spark-submit, spark-defaults.conf etc here and there. Author: Andrew Or <andrewor14@gmail.com> Closes #701 from andrewor14/yarn-docs and squashes the following commits: e2c2312 [Andrew Or] Merge in changes in #752 (SPARK-1814) 25cfe7b [Andrew Or] Merge in the warning from SPARK-1753 a8c39c5 [Andrew Or] Minor changes 336bbd9 [Andrew Or] Tabs -> spaces 4d9d8f7 [Andrew Or] Merge branch 'master' of github.com:apache/spark into yarn-docs 041017a [Andrew Or] Abstract Spark submit documentation to cluster-overview.html 3cc0649 [Andrew Or] Detail how to set configurations + remove legacy instructions 5b7140a [Andrew Or] Merge branch 'master' of github.com:apache/spark into yarn-docs 85a51fc [Andrew Or] Update run-example, spark-shell, configuration etc. c10e8c7 [Andrew Or] Merge branch 'master' of github.com:apache/spark into yarn-docs 381fe32 [Andrew Or] Update docs for standalone mode 757c184 [Andrew Or] Add a note about the requirements for the debugging trick f8ca990 [Andrew Or] Merge branch 'master' of github.com:apache/spark into yarn-docs 924f04c [Andrew Or] Revert addition of --deploy-mode d5fe17b [Andrew Or] Update the YARN docs
4.5 KiB
layout | title |
---|---|
global | Running with Cloudera and HortonWorks |
Spark can run against all versions of Cloudera's Distribution Including Apache Hadoop (CDH) and the Hortonworks Data Platform (HDP). There are a few things to keep in mind when using Spark with these distributions:
Compile-time Hadoop Version
When compiling Spark, you'll need to specify the Hadoop version by defining the hadoop.version
property. For certain versions, you will need to specify additional profiles. For more detail,
see the guide on building with maven:
mvn -Dhadoop.version=1.0.4 -DskipTests clean package
mvn -Phadoop-2.2 -Dhadoop.version=2.2.0 -DskipTests clean package
The table below lists the corresponding hadoop.version
code for each CDH/HDP release. Note that
some Hadoop releases are binary compatible across client versions. This means the pre-built Spark
distribution may "just work" without you needing to compile. That said, we recommend compiling with
the exact Hadoop version you are running to avoid any compatibility errors.
CDH Releases
|
HDP Releases
|
In SBT, the equivalent can be achieved by setting the SPARK_HADOOP_VERSION flag:
SPARK_HADOOP_VERSION=1.0.4 sbt/sbt assembly
Linking Applications to the Hadoop Version
In addition to compiling Spark itself against the right version, you need to add a Maven dependency on that
version of hadoop-client
to any Spark applications you run, so they can also talk to the HDFS version
on the cluster. If you are using CDH, you also need to add the Cloudera Maven repository.
This looks as follows in SBT:
{% highlight scala %} libraryDependencies += "org.apache.hadoop" % "hadoop-client" % ""
// If using CDH, also add Cloudera repo resolvers += "Cloudera Repository" at "https://repository.cloudera.com/artifactory/cloudera-repos/" {% endhighlight %}
Or in Maven:
{% highlight xml %} ... org.apache.hadoop hadoop-client [version]
... Cloudera repository https://repository.cloudera.com/artifactory/cloudera-repos/{% endhighlight %}
Where to Run Spark
As described in the Hardware Provisioning guide, Spark can run in a variety of deployment modes:
- Using dedicated set of Spark nodes in your cluster. These nodes should be co-located with your Hadoop installation.
- Running on the same nodes as an existing Hadoop installation, with a fixed amount memory and cores dedicated to Spark on each node.
- Run Spark alongside Hadoop using a cluster resource manager, such as YARN or Mesos.
These options are identical for those using CDH and HDP.
Inheriting Cluster Configuration
If you plan to read and write from HDFS using Spark, there are two Hadoop configuration files that should be included on Spark's classpath:
hdfs-site.xml
, which provides default behaviors for the HDFS client.core-site.xml
, which sets the default filesystem name.
The location of these configuration files varies across CDH and HDP versions, but
a common location is inside of /etc/hadoop/conf
. Some tools, such as Cloudera Manager, create
configurations on-the-fly, but offer a mechanisms to download copies of them.
To make these files visible to Spark, set HADOOP_CONF_DIR
in $SPARK_HOME/spark-env.sh
to a location containing the configuration files.