| |
| <!DOCTYPE html> |
| <!--[if lt IE 7]> <html class="no-js lt-ie9 lt-ie8 lt-ie7"> <![endif]--> |
| <!--[if IE 7]> <html class="no-js lt-ie9 lt-ie8"> <![endif]--> |
| <!--[if IE 8]> <html class="no-js lt-ie9"> <![endif]--> |
| <!--[if gt IE 8]><!--> <html class="no-js"> <!--<![endif]--> |
| <head> |
| <meta charset="utf-8"> |
| <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> |
| <title>Running Spark on Mesos - Spark 2.2.2 Documentation</title> |
| |
| |
| |
| |
| <link rel="stylesheet" href="css/bootstrap.min.css"> |
| <style> |
| body { |
| padding-top: 60px; |
| padding-bottom: 40px; |
| } |
| </style> |
| <meta name="viewport" content="width=device-width"> |
| <link rel="stylesheet" href="css/bootstrap-responsive.min.css"> |
| <link rel="stylesheet" href="css/main.css"> |
| |
| <script src="js/vendor/modernizr-2.6.1-respond-1.1.0.min.js"></script> |
| |
| <link rel="stylesheet" href="css/pygments-default.css"> |
| |
| |
| <!-- Google analytics script --> |
| <script type="text/javascript"> |
| var _gaq = _gaq || []; |
| _gaq.push(['_setAccount', 'UA-32518208-2']); |
| _gaq.push(['_trackPageview']); |
| |
| (function() { |
| var ga = document.createElement('script'); ga.type = 'text/javascript'; ga.async = true; |
| ga.src = ('https:' == document.location.protocol ? 'https://ssl' : 'http://www') + '.google-analytics.com/ga.js'; |
| var s = document.getElementsByTagName('script')[0]; s.parentNode.insertBefore(ga, s); |
| })(); |
| </script> |
| |
| |
| </head> |
| <body> |
| <!--[if lt IE 7]> |
| <p class="chromeframe">You are using an outdated browser. <a href="http://browsehappy.com/">Upgrade your browser today</a> or <a href="http://www.google.com/chromeframe/?redirect=true">install Google Chrome Frame</a> to better experience this site.</p> |
| <![endif]--> |
| |
| <!-- This code is taken from http://twitter.github.com/bootstrap/examples/hero.html --> |
| |
| <div class="navbar navbar-fixed-top" id="topbar"> |
| <div class="navbar-inner"> |
| <div class="container"> |
| <div class="brand"><a href="index.html"> |
| <img src="img/spark-logo-hd.png" style="height:50px;"/></a><span class="version">2.2.2</span> |
| </div> |
| <ul class="nav"> |
| <!--TODO(andyk): Add class="active" attribute to li some how.--> |
| <li><a href="index.html">Overview</a></li> |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">Programming Guides<b class="caret"></b></a> |
| <ul class="dropdown-menu"> |
| <li><a href="quick-start.html">Quick Start</a></li> |
| <li><a href="rdd-programming-guide.html">RDDs, Accumulators, Broadcasts Vars</a></li> |
| <li><a href="sql-programming-guide.html">SQL, DataFrames, and Datasets</a></li> |
| <li><a href="structured-streaming-programming-guide.html">Structured Streaming</a></li> |
| <li><a href="streaming-programming-guide.html">Spark Streaming (DStreams)</a></li> |
| <li><a href="ml-guide.html">MLlib (Machine Learning)</a></li> |
| <li><a href="graphx-programming-guide.html">GraphX (Graph Processing)</a></li> |
| <li><a href="sparkr.html">SparkR (R on Spark)</a></li> |
| </ul> |
| </li> |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">API Docs<b class="caret"></b></a> |
| <ul class="dropdown-menu"> |
| <li><a href="api/scala/index.html#org.apache.spark.package">Scala</a></li> |
| <li><a href="api/java/index.html">Java</a></li> |
| <li><a href="api/python/index.html">Python</a></li> |
| <li><a href="api/R/index.html">R</a></li> |
| </ul> |
| </li> |
| |
| <li class="dropdown"> |
| <a href="#" class="dropdown-toggle" data-toggle="dropdown">Deploying<b class="caret"></b></a> |
| <ul class="dropdown-menu"> |
| <li><a href="cluster-overview.html">Overview</a></li> |
| <li><a href="submitting-applications.html">Submitting Applications</a></li> |
| <li class="divider"></li> |
| <li><a href="spark-standalone.html">Spark Standalone</a></li> |
| <li><a href="running-on-mesos.html">Mesos</a></li> |
| <li><a href="running-on-yarn.html">YARN</a></li> |
| </ul> |
| </li> |
| |
| <li class="dropdown"> |
| <a href="api.html" class="dropdown-toggle" data-toggle="dropdown">More<b class="caret"></b></a> |
| <ul class="dropdown-menu"> |
| <li><a href="configuration.html">Configuration</a></li> |
| <li><a href="monitoring.html">Monitoring</a></li> |
| <li><a href="tuning.html">Tuning Guide</a></li> |
| <li><a href="job-scheduling.html">Job Scheduling</a></li> |
| <li><a href="security.html">Security</a></li> |
| <li><a href="hardware-provisioning.html">Hardware Provisioning</a></li> |
| <li class="divider"></li> |
| <li><a href="building-spark.html">Building Spark</a></li> |
| <li><a href="http://spark.apache.org/contributing.html">Contributing to Spark</a></li> |
| <li><a href="http://spark.apache.org/third-party-projects.html">Third Party Projects</a></li> |
| </ul> |
| </li> |
| </ul> |
| <!--<p class="navbar-text pull-right"><span class="version-text">v2.2.2</span></p>--> |
| </div> |
| </div> |
| </div> |
| |
| <div class="container-wrapper"> |
| |
| |
| <div class="content" id="content"> |
| |
| <h1 class="title">Running Spark on Mesos</h1> |
| |
| |
| <ul id="markdown-toc"> |
| <li><a href="#how-it-works" id="markdown-toc-how-it-works">How it Works</a></li> |
| <li><a href="#installing-mesos" id="markdown-toc-installing-mesos">Installing Mesos</a> <ul> |
| <li><a href="#from-source" id="markdown-toc-from-source">From Source</a></li> |
| <li><a href="#third-party-packages" id="markdown-toc-third-party-packages">Third-Party Packages</a></li> |
| <li><a href="#verification" id="markdown-toc-verification">Verification</a></li> |
| </ul> |
| </li> |
| <li><a href="#connecting-spark-to-mesos" id="markdown-toc-connecting-spark-to-mesos">Connecting Spark to Mesos</a> <ul> |
| <li><a href="#uploading-spark-package" id="markdown-toc-uploading-spark-package">Uploading Spark Package</a></li> |
| <li><a href="#using-a-mesos-master-url" id="markdown-toc-using-a-mesos-master-url">Using a Mesos Master URL</a></li> |
| <li><a href="#client-mode" id="markdown-toc-client-mode">Client Mode</a></li> |
| <li><a href="#cluster-mode" id="markdown-toc-cluster-mode">Cluster mode</a></li> |
| </ul> |
| </li> |
| <li><a href="#mesos-run-modes" id="markdown-toc-mesos-run-modes">Mesos Run Modes</a> <ul> |
| <li><a href="#coarse-grained" id="markdown-toc-coarse-grained">Coarse-Grained</a></li> |
| <li><a href="#fine-grained-deprecated" id="markdown-toc-fine-grained-deprecated">Fine-Grained (deprecated)</a></li> |
| </ul> |
| </li> |
| <li><a href="#mesos-docker-support" id="markdown-toc-mesos-docker-support">Mesos Docker Support</a></li> |
| <li><a href="#running-alongside-hadoop" id="markdown-toc-running-alongside-hadoop">Running Alongside Hadoop</a></li> |
| <li><a href="#dynamic-resource-allocation-with-mesos" id="markdown-toc-dynamic-resource-allocation-with-mesos">Dynamic Resource Allocation with Mesos</a></li> |
| <li><a href="#configuration" id="markdown-toc-configuration">Configuration</a> <ul> |
| <li><a href="#spark-properties" id="markdown-toc-spark-properties">Spark Properties</a></li> |
| </ul> |
| </li> |
| <li><a href="#troubleshooting-and-debugging" id="markdown-toc-troubleshooting-and-debugging">Troubleshooting and Debugging</a></li> |
| </ul> |
| |
| <p>Spark can run on hardware clusters managed by <a href="http://mesos.apache.org/">Apache Mesos</a>.</p> |
| |
| <p>The advantages of deploying Spark with Mesos include:</p> |
| |
| <ul> |
| <li>dynamic partitioning between Spark and other |
| <a href="https://mesos.apache.org/documentation/latest/mesos-frameworks/">frameworks</a></li> |
| <li>scalable partitioning between multiple instances of Spark</li> |
| </ul> |
| |
| <h1 id="how-it-works">How it Works</h1> |
| |
| <p>In a standalone cluster deployment, the cluster manager in the below diagram is a Spark master |
| instance. When using Mesos, the Mesos master replaces the Spark master as the cluster manager.</p> |
| |
| <p style="text-align: center;"> |
| <img src="img/cluster-overview.png" title="Spark cluster components" alt="Spark cluster components" /> |
| </p> |
| |
| <p>Now when a driver creates a job and starts issuing tasks for scheduling, Mesos determines what |
| machines handle what tasks. Because it takes into account other frameworks when scheduling these |
| many short-lived tasks, multiple frameworks can coexist on the same cluster without resorting to a |
| static partitioning of resources.</p> |
| |
| <p>To get started, follow the steps below to install Mesos and deploy Spark jobs via Mesos.</p> |
| |
| <h1 id="installing-mesos">Installing Mesos</h1> |
| |
| <p>Spark 2.2.2 is designed for use with Mesos 1.0.0 or newer and does not |
| require any special patches of Mesos.</p> |
| |
| <p>If you already have a Mesos cluster running, you can skip this Mesos installation step.</p> |
| |
| <p>Otherwise, installing Mesos for Spark is no different than installing Mesos for use by other |
| frameworks. You can install Mesos either from source or using prebuilt packages.</p> |
| |
| <h2 id="from-source">From Source</h2> |
| |
| <p>To install Apache Mesos from source, follow these steps:</p> |
| |
| <ol> |
| <li>Download a Mesos release from a |
| <a href="http://www.apache.org/dyn/closer.lua/mesos/1.0.0/">mirror</a></li> |
| <li>Follow the Mesos <a href="http://mesos.apache.org/gettingstarted">Getting Started</a> page for compiling and |
| installing Mesos</li> |
| </ol> |
| |
| <p><strong>Note:</strong> If you want to run Mesos without installing it into the default paths on your system |
| (e.g., if you lack administrative privileges to install it), pass the |
| <code>--prefix</code> option to <code>configure</code> to tell it where to install. For example, pass |
| <code>--prefix=/home/me/mesos</code>. By default the prefix is <code>/usr/local</code>.</p> |
| |
| <h2 id="third-party-packages">Third-Party Packages</h2> |
| |
| <p>The Apache Mesos project only publishes source releases, not binary packages. But other |
| third party projects publish binary releases that may be helpful in setting Mesos up.</p> |
| |
| <p>One of those is Mesosphere. To install Mesos using the binary releases provided by Mesosphere:</p> |
| |
| <ol> |
| <li>Download Mesos installation package from <a href="http://mesosphere.io/downloads/">downloads page</a></li> |
| <li>Follow their instructions for installation and configuration</li> |
| </ol> |
| |
| <p>The Mesosphere installation documents suggest setting up ZooKeeper to handle Mesos master failover, |
| but Mesos can be run without ZooKeeper using a single master as well.</p> |
| |
| <h2 id="verification">Verification</h2> |
| |
| <p>To verify that the Mesos cluster is ready for Spark, navigate to the Mesos master webui at port |
| <code>:5050</code> Confirm that all expected machines are present in the slaves tab.</p> |
| |
| <h1 id="connecting-spark-to-mesos">Connecting Spark to Mesos</h1> |
| |
| <p>To use Mesos from Spark, you need a Spark binary package available in a place accessible by Mesos, and |
| a Spark driver program configured to connect to Mesos.</p> |
| |
| <p>Alternatively, you can also install Spark in the same location in all the Mesos slaves, and configure |
| <code>spark.mesos.executor.home</code> (defaults to SPARK_HOME) to point to that location.</p> |
| |
| <h2 id="uploading-spark-package">Uploading Spark Package</h2> |
| |
| <p>When Mesos runs a task on a Mesos slave for the first time, that slave must have a Spark binary |
| package for running the Spark Mesos executor backend. |
| The Spark package can be hosted at any Hadoop-accessible URI, including HTTP via <code>http://</code>, |
| <a href="http://aws.amazon.com/s3">Amazon Simple Storage Service</a> via <code>s3n://</code>, or HDFS via <code>hdfs://</code>.</p> |
| |
| <p>To use a precompiled package:</p> |
| |
| <ol> |
| <li>Download a Spark binary package from the Spark <a href="https://spark.apache.org/downloads.html">download page</a></li> |
| <li>Upload to hdfs/http/s3</li> |
| </ol> |
| |
| <p>To host on HDFS, use the Hadoop fs put command: <code>hadoop fs -put spark-2.2.2.tar.gz |
| /path/to/spark-2.2.2.tar.gz</code></p> |
| |
| <p>Or if you are using a custom-compiled version of Spark, you will need to create a package using |
| the <code>dev/make-distribution.sh</code> script included in a Spark source tarball/checkout.</p> |
| |
| <ol> |
| <li>Download and build Spark using the instructions <a href="index.html">here</a></li> |
| <li>Create a binary package using <code>./dev/make-distribution.sh --tgz</code>.</li> |
| <li>Upload archive to http/s3/hdfs</li> |
| </ol> |
| |
| <h2 id="using-a-mesos-master-url">Using a Mesos Master URL</h2> |
| |
| <p>The Master URLs for Mesos are in the form <code>mesos://host:5050</code> for a single-master Mesos |
| cluster, or <code>mesos://zk://host1:2181,host2:2181,host3:2181/mesos</code> for a multi-master Mesos cluster using ZooKeeper.</p> |
| |
| <h2 id="client-mode">Client Mode</h2> |
| |
| <p>In client mode, a Spark Mesos framework is launched directly on the client machine and waits for the driver output.</p> |
| |
| <p>The driver needs some configuration in <code>spark-env.sh</code> to interact properly with Mesos:</p> |
| |
| <ol> |
| <li>In <code>spark-env.sh</code> set some environment variables: |
| <ul> |
| <li><code>export MESOS_NATIVE_JAVA_LIBRARY=<path to libmesos.so></code>. This path is typically |
| <code><prefix>/lib/libmesos.so</code> where the prefix is <code>/usr/local</code> by default. See Mesos installation |
| instructions above. On Mac OS X, the library is called <code>libmesos.dylib</code> instead of |
| <code>libmesos.so</code>.</li> |
| <li><code>export SPARK_EXECUTOR_URI=<URL of spark-2.2.2.tar.gz uploaded above></code>.</li> |
| </ul> |
| </li> |
| <li>Also set <code>spark.executor.uri</code> to <code><URL of spark-2.2.2.tar.gz></code>.</li> |
| </ol> |
| |
| <p>Now when starting a Spark application against the cluster, pass a <code>mesos://</code> |
| URL as the master when creating a <code>SparkContext</code>. For example:</p> |
| |
| <figure class="highlight"><pre><code class="language-scala" data-lang="scala"><span></span><span class="k">val</span> <span class="n">conf</span> <span class="k">=</span> <span class="k">new</span> <span class="nc">SparkConf</span><span class="o">()</span> |
| <span class="o">.</span><span class="n">setMaster</span><span class="o">(</span><span class="s">"mesos://HOST:5050"</span><span class="o">)</span> |
| <span class="o">.</span><span class="n">setAppName</span><span class="o">(</span><span class="s">"My app"</span><span class="o">)</span> |
| <span class="o">.</span><span class="n">set</span><span class="o">(</span><span class="s">"spark.executor.uri"</span><span class="o">,</span> <span class="s">"<path to spark-2.2.2.tar.gz uploaded above>"</span><span class="o">)</span> |
| <span class="k">val</span> <span class="n">sc</span> <span class="k">=</span> <span class="k">new</span> <span class="nc">SparkContext</span><span class="o">(</span><span class="n">conf</span><span class="o">)</span></code></pre></figure> |
| |
| <p>(You can also use <a href="submitting-applications.html"><code>spark-submit</code></a> and configure <code>spark.executor.uri</code> |
| in the <a href="configuration.html#loading-default-configurations">conf/spark-defaults.conf</a> file.)</p> |
| |
| <p>When running a shell, the <code>spark.executor.uri</code> parameter is inherited from <code>SPARK_EXECUTOR_URI</code>, so |
| it does not need to be redundantly passed in as a system property.</p> |
| |
| <figure class="highlight"><pre><code class="language-bash" data-lang="bash"><span></span>./bin/spark-shell --master mesos://host:5050</code></pre></figure> |
| |
| <h2 id="cluster-mode">Cluster mode</h2> |
| |
| <p>Spark on Mesos also supports cluster mode, where the driver is launched in the cluster and the client |
| can find the results of the driver from the Mesos Web UI.</p> |
| |
| <p>To use cluster mode, you must start the <code>MesosClusterDispatcher</code> in your cluster via the <code>sbin/start-mesos-dispatcher.sh</code> script, |
| passing in the Mesos master URL (e.g: mesos://host:5050). This starts the <code>MesosClusterDispatcher</code> as a daemon running on the host.</p> |
| |
| <p>If you like to run the <code>MesosClusterDispatcher</code> with Marathon, you need to run the <code>MesosClusterDispatcher</code> in the foreground (i.e: <code>bin/spark-class org.apache.spark.deploy.mesos.MesosClusterDispatcher</code>). Note that the <code>MesosClusterDispatcher</code> not yet supports multiple instances for HA.</p> |
| |
| <p>The <code>MesosClusterDispatcher</code> also supports writing recovery state into Zookeeper. This will allow the <code>MesosClusterDispatcher</code> to be able to recover all submitted and running containers on relaunch. In order to enable this recovery mode, you can set SPARK_DAEMON_JAVA_OPTS in spark-env by configuring <code>spark.deploy.recoveryMode</code> and related spark.deploy.zookeeper.* configurations. |
| For more information about these configurations please refer to the configurations <a href="configurations.html#deploy">doc</a>.</p> |
| |
| <p>You can also specify any additional jars required by the <code>MesosClusterDispatcher</code> in the classpath by setting the environment variable SPARK_DAEMON_CLASSPATH in spark-env.</p> |
| |
| <p>From the client, you can submit a job to Mesos cluster by running <code>spark-submit</code> and specifying the master URL |
| to the URL of the <code>MesosClusterDispatcher</code> (e.g: mesos://dispatcher:7077). You can view driver statuses on the |
| Spark cluster Web UI.</p> |
| |
| <p>For example:</p> |
| |
| <figure class="highlight"><pre><code class="language-bash" data-lang="bash"><span></span>./bin/spark-submit <span class="se">\</span> |
| --class org.apache.spark.examples.SparkPi <span class="se">\</span> |
| --master mesos://207.184.161.138:7077 <span class="se">\</span> |
| --deploy-mode cluster <span class="se">\</span> |
| --supervise <span class="se">\</span> |
| --executor-memory 20G <span class="se">\</span> |
| --total-executor-cores <span class="m">100</span> <span class="se">\</span> |
| http://path/to/examples.jar <span class="se">\</span> |
| <span class="m">1000</span></code></pre></figure> |
| |
| <p>Note that jars or python files that are passed to spark-submit should be URIs reachable by Mesos slaves, as the Spark driver doesn’t automatically upload local jars.</p> |
| |
| <h1 id="mesos-run-modes">Mesos Run Modes</h1> |
| |
| <p>Spark can run over Mesos in two modes: “coarse-grained” (default) and |
| “fine-grained” (deprecated).</p> |
| |
| <h2 id="coarse-grained">Coarse-Grained</h2> |
| |
| <p>In “coarse-grained” mode, each Spark executor runs as a single Mesos |
| task. Spark executors are sized according to the following |
| configuration variables:</p> |
| |
| <ul> |
| <li>Executor memory: <code>spark.executor.memory</code></li> |
| <li>Executor cores: <code>spark.executor.cores</code></li> |
| <li>Number of executors: <code>spark.cores.max</code>/<code>spark.executor.cores</code></li> |
| </ul> |
| |
| <p>Please see the <a href="configuration.html">Spark Configuration</a> page for |
| details and default values.</p> |
| |
| <p>Executors are brought up eagerly when the application starts, until |
| <code>spark.cores.max</code> is reached. If you don’t set <code>spark.cores.max</code>, the |
| Spark application will reserve all resources offered to it by Mesos, |
| so we of course urge you to set this variable in any sort of |
| multi-tenant cluster, including one which runs multiple concurrent |
| Spark applications.</p> |
| |
| <p>The scheduler will start executors round-robin on the offers Mesos |
| gives it, but there are no spread guarantees, as Mesos does not |
| provide such guarantees on the offer stream.</p> |
| |
| <p>In this mode spark executors will honor port allocation if such is |
| provided from the user. Specifically if the user defines |
| <code>spark.executor.port</code> or <code>spark.blockManager.port</code> in Spark configuration, |
| the mesos scheduler will check the available offers for a valid port |
| range containing the port numbers. If no such range is available it will |
| not launch any task. If no restriction is imposed on port numbers by the |
| user, ephemeral ports are used as usual. This port honouring implementation |
| implies one task per host if the user defines a port. In the future network |
| isolation shall be supported.</p> |
| |
| <p>The benefit of coarse-grained mode is much lower startup overhead, but |
| at the cost of reserving Mesos resources for the complete duration of |
| the application. To configure your job to dynamically adjust to its |
| resource requirements, look into |
| <a href="#dynamic-resource-allocation-with-mesos">Dynamic Allocation</a>.</p> |
| |
| <h2 id="fine-grained-deprecated">Fine-Grained (deprecated)</h2> |
| |
| <p><strong>NOTE:</strong> Fine-grained mode is deprecated as of Spark 2.0.0. Consider |
| using <a href="#dynamic-resource-allocation-with-mesos">Dynamic Allocation</a> |
| for some of the benefits. For a full explanation see |
| <a href="https://issues.apache.org/jira/browse/SPARK-11857">SPARK-11857</a></p> |
| |
| <p>In “fine-grained” mode, each Spark task inside the Spark executor runs |
| as a separate Mesos task. This allows multiple instances of Spark (and |
| other frameworks) to share cores at a very fine granularity, where |
| each application gets more or fewer cores as it ramps up and down, but |
| it comes with an additional overhead in launching each task. This mode |
| may be inappropriate for low-latency requirements like interactive |
| queries or serving web requests.</p> |
| |
| <p>Note that while Spark tasks in fine-grained will relinquish cores as |
| they terminate, they will not relinquish memory, as the JVM does not |
| give memory back to the Operating System. Neither will executors |
| terminate when they’re idle.</p> |
| |
| <p>To run in fine-grained mode, set the <code>spark.mesos.coarse</code> property to false in your |
| <a href="configuration.html#spark-properties">SparkConf</a>:</p> |
| |
| <figure class="highlight"><pre><code class="language-scala" data-lang="scala"><span></span><span class="n">conf</span><span class="o">.</span><span class="n">set</span><span class="o">(</span><span class="s">"spark.mesos.coarse"</span><span class="o">,</span> <span class="s">"false"</span><span class="o">)</span></code></pre></figure> |
| |
| <p>You may also make use of <code>spark.mesos.constraints</code> to set |
| attribute-based constraints on Mesos resource offers. By default, all |
| resource offers will be accepted.</p> |
| |
| <figure class="highlight"><pre><code class="language-scala" data-lang="scala"><span></span><span class="n">conf</span><span class="o">.</span><span class="n">set</span><span class="o">(</span><span class="s">"spark.mesos.constraints"</span><span class="o">,</span> <span class="s">"os:centos7;us-east-1:false"</span><span class="o">)</span></code></pre></figure> |
| |
| <p>For example, Let’s say <code>spark.mesos.constraints</code> is set to <code>os:centos7;us-east-1:false</code>, then the resource offers will |
| be checked to see if they meet both these constraints and only then will be accepted to start new executors.</p> |
| |
| <p>To constrain where driver tasks are run, use <code>spark.mesos.driver.constraints</code></p> |
| |
| <h1 id="mesos-docker-support">Mesos Docker Support</h1> |
| |
| <p>Spark can make use of a Mesos Docker containerizer by setting the property <code>spark.mesos.executor.docker.image</code> |
| in your <a href="configuration.html#spark-properties">SparkConf</a>.</p> |
| |
| <p>The Docker image used must have an appropriate version of Spark already part of the image, or you can |
| have Mesos download Spark via the usual methods.</p> |
| |
| <p>Requires Mesos version 0.20.1 or later.</p> |
| |
| <p>Note that by default Mesos agents will not pull the image if it already exists on the agent. If you use mutable image |
| tags you can set <code>spark.mesos.executor.docker.forcePullImage</code> to <code>true</code> in order to force the agent to always pull the |
| image before running the executor. Force pulling images is only available in Mesos version 0.22 and above.</p> |
| |
| <h1 id="running-alongside-hadoop">Running Alongside Hadoop</h1> |
| |
| <p>You can run Spark and Mesos alongside your existing Hadoop cluster by just launching them as a |
| separate service on the machines. To access Hadoop data from Spark, a full <code>hdfs://</code> URL is required |
| (typically <code>hdfs://<namenode>:9000/path</code>, but you can find the right URL on your Hadoop Namenode web |
| UI).</p> |
| |
| <p>In addition, it is possible to also run Hadoop MapReduce on Mesos for better resource isolation and |
| sharing between the two. In this case, Mesos will act as a unified scheduler that assigns cores to |
| either Hadoop or Spark, as opposed to having them share resources via the Linux scheduler on each |
| node. Please refer to <a href="https://github.com/mesos/hadoop">Hadoop on Mesos</a>.</p> |
| |
| <p>In either case, HDFS runs separately from Hadoop MapReduce, without being scheduled through Mesos.</p> |
| |
| <h1 id="dynamic-resource-allocation-with-mesos">Dynamic Resource Allocation with Mesos</h1> |
| |
| <p>Mesos supports dynamic allocation only with coarse-grained mode, which can resize the number of |
| executors based on statistics of the application. For general information, |
| see <a href="job-scheduling.html#dynamic-resource-allocation">Dynamic Resource Allocation</a>.</p> |
| |
| <p>The External Shuffle Service to use is the Mesos Shuffle Service. It provides shuffle data cleanup functionality |
| on top of the Shuffle Service since Mesos doesn’t yet support notifying another framework’s |
| termination. To launch it, run <code>$SPARK_HOME/sbin/start-mesos-shuffle-service.sh</code> on all slave nodes, with <code>spark.shuffle.service.enabled</code> set to <code>true</code>.</p> |
| |
| <p>This can also be achieved through Marathon, using a unique host constraint, and the following command: <code>bin/spark-class org.apache.spark.deploy.mesos.MesosExternalShuffleService</code>.</p> |
| |
| <h1 id="configuration">Configuration</h1> |
| |
| <p>See the <a href="configuration.html">configuration page</a> for information on Spark configurations. The following configs are specific for Spark on Mesos.</p> |
| |
| <h4 id="spark-properties">Spark Properties</h4> |
| |
| <table class="table"> |
| <tr><th>Property Name</th><th>Default</th><th>Meaning</th></tr> |
| <tr> |
| <td><code>spark.mesos.coarse</code></td> |
| <td>true</td> |
| <td> |
| If set to <code>true</code>, runs over Mesos clusters in "coarse-grained" sharing mode, where Spark acquires one long-lived Mesos task on each machine. |
| If set to <code>false</code>, runs over Mesos cluster in "fine-grained" sharing mode, where one Mesos task is created per Spark task. |
| Detailed information in <a href="running-on-mesos.html#mesos-run-modes">'Mesos Run Modes'</a>. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.extra.cores</code></td> |
| <td><code>0</code></td> |
| <td> |
| Set the extra number of cores for an executor to advertise. This |
| does not result in more cores allocated. It instead means that an |
| executor will "pretend" it has more cores, so that the driver will |
| send it more tasks. Use this to increase parallelism. This |
| setting is only used for Mesos coarse-grained mode. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.mesosExecutor.cores</code></td> |
| <td><code>1.0</code></td> |
| <td> |
| (Fine-grained mode only) Number of cores to give each Mesos executor. This does not |
| include the cores used to run the Spark tasks. In other words, even if no Spark task |
| is being run, each Mesos executor will occupy the number of cores configured here. |
| The value can be a floating point number. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.docker.image</code></td> |
| <td>(none)</td> |
| <td> |
| Set the name of the docker image that the Spark executors will run in. The selected |
| image must have Spark installed, as well as a compatible version of the Mesos library. |
| The installed path of Spark in the image can be specified with <code>spark.mesos.executor.home</code>; |
| the installed path of the Mesos library can be specified with <code>spark.executorEnv.MESOS_NATIVE_JAVA_LIBRARY</code>. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.docker.forcePullImage</code></td> |
| <td>false</td> |
| <td> |
| Force Mesos agents to pull the image specified in <code>spark.mesos.executor.docker.image</code>. |
| By default Mesos agents will not pull images they already have cached. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.docker.parameters</code></td> |
| <td>(none)</td> |
| <td> |
| Set the list of custom parameters which will be passed into the <code>docker run</code> command when launching the Spark executor on Mesos using the docker containerizer. The format of this property is a comma-separated list of |
| key/value pairs. Example: |
| |
| <pre>key1=val1,key2=val2,key3=val3</pre> |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.docker.volumes</code></td> |
| <td>(none)</td> |
| <td> |
| Set the list of volumes which will be mounted into the Docker image, which was set using |
| <code>spark.mesos.executor.docker.image</code>. The format of this property is a comma-separated list of |
| mappings following the form passed to <code>docker run -v</code>. That is they take the form: |
| |
| <pre>[host_path:]container_path[:ro|:rw]</pre> |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.task.labels</code></td> |
| <td>(none)</td> |
| <td> |
| Set the Mesos labels to add to each task. Labels are free-form key-value pairs. |
| Key-value pairs should be separated by a colon, and commas used to list more than one. |
| Ex. key:value,key2:value2. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.home</code></td> |
| <td>driver side <code>SPARK_HOME</code></td> |
| <td> |
| Set the directory in which Spark is installed on the executors in Mesos. By default, the |
| executors will simply use the driver's Spark home directory, which may not be visible to |
| them. Note that this is only relevant if a Spark binary package is not specified through |
| <code>spark.executor.uri</code>. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.executor.memoryOverhead</code></td> |
| <td>executor memory * 0.10, with minimum of 384</td> |
| <td> |
| The amount of additional memory, specified in MB, to be allocated per executor. By default, |
| the overhead will be larger of either 384 or 10% of <code>spark.executor.memory</code>. If set, |
| the final overhead will be this value. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.uris</code></td> |
| <td>(none)</td> |
| <td> |
| A comma-separated list of URIs to be downloaded to the sandbox |
| when driver or executor is launched by Mesos. This applies to |
| both coarse-grained and fine-grained mode. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.principal</code></td> |
| <td>(none)</td> |
| <td> |
| Set the principal with which Spark framework will use to authenticate with Mesos. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.secret</code></td> |
| <td>(none)</td> |
| <td> |
| Set the secret with which Spark framework will use to authenticate with Mesos. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.role</code></td> |
| <td><code>*</code></td> |
| <td> |
| Set the role of this Spark framework for Mesos. Roles are used in Mesos for reservations |
| and resource weight sharing. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.constraints</code></td> |
| <td>(none)</td> |
| <td> |
| Attribute based constraints on mesos resource offers. By default, all resource offers will be accepted. This setting |
| applies only to executors. Refer to <a href="http://mesos.apache.org/documentation/attributes-resources/">Mesos |
| Attributes & Resources</a> for more information on attributes. |
| <ul> |
| <li>Scalar constraints are matched with "less than equal" semantics i.e. value in the constraint must be less than or equal to the value in the resource offer.</li> |
| <li>Range constraints are matched with "contains" semantics i.e. value in the constraint must be within the resource offer's value.</li> |
| <li>Set constraints are matched with "subset of" semantics i.e. value in the constraint must be a subset of the resource offer's value.</li> |
| <li>Text constraints are matched with "equality" semantics i.e. value in the constraint must be exactly equal to the resource offer's value.</li> |
| <li>In case there is no value present as a part of the constraint any offer with the corresponding attribute will be accepted (without value check).</li> |
| </ul> |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.driver.constraints</code></td> |
| <td>(none)</td> |
| <td> |
| Same as <code>spark.mesos.constraints</code> except applied to drivers when launched through the dispatcher. By default, |
| all offers with sufficient resources will be accepted. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.containerizer</code></td> |
| <td><code>docker</code></td> |
| <td> |
| This only affects docker containers, and must be one of "docker" |
| or "mesos". Mesos supports two types of |
| containerizers for docker: the "docker" containerizer, and the preferred |
| "mesos" containerizer. Read more here: http://mesos.apache.org/documentation/latest/container-image/ |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.driver.webui.url</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| Set the Spark Mesos driver webui_url for interacting with the framework. |
| If unset it will point to Spark's internal web UI. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.driverEnv.[EnvironmentVariableName]</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| This only affects drivers submitted in cluster mode. Add the |
| environment variable specified by EnvironmentVariableName to the |
| driver process. The user can specify multiple of these to set |
| multiple environment variables. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.dispatcher.webui.url</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| Set the Spark Mesos dispatcher webui_url for interacting with the framework. |
| If unset it will point to Spark's internal web UI. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.dispatcher.driverDefault.[PropertyName]</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| Set default properties for drivers submitted through the |
| dispatcher. For example, |
| spark.mesos.dispatcher.driverProperty.spark.executor.memory=32g |
| results in the executors for all drivers submitted in cluster mode |
| to run in 32g containers. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.dispatcher.historyServer.url</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| Set the URL of the <a href="http://spark.apache.org/docs/latest/monitoring.html#viewing-after-the-fact">history |
| server</a>. The dispatcher will then link each driver to its entry |
| in the history server. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.gpus.max</code></td> |
| <td><code>0</code></td> |
| <td> |
| Set the maximum number GPU resources to acquire for this job. Note that executors will still launch when no GPU resources are found |
| since this configuration is just a upper limit and not a guaranteed amount. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.network.name</code></td> |
| <td><code>(none)</code></td> |
| <td> |
| Attach containers to the given named network. If this job is |
| launched in cluster mode, also launch the driver in the given named |
| network. See |
| <a href="http://mesos.apache.org/documentation/latest/cni/">the Mesos CNI docs</a> |
| for more details. |
| </td> |
| </tr> |
| <tr> |
| <td><code>spark.mesos.fetcherCache.enable</code></td> |
| <td><code>false</code></td> |
| <td> |
| If set to `true`, all URIs (example: `spark.executor.uri`, |
| `spark.mesos.uris`) will be cached by the <a href="http://mesos.apache.org/documentation/latest/fetcher/">Mesos |
| Fetcher Cache</a> |
| </td> |
| </tr> |
| </table> |
| |
| <h1 id="troubleshooting-and-debugging">Troubleshooting and Debugging</h1> |
| |
| <p>A few places to look during debugging:</p> |
| |
| <ul> |
| <li>Mesos master on port <code>:5050</code> |
| <ul> |
| <li>Slaves should appear in the slaves tab</li> |
| <li>Spark applications should appear in the frameworks tab</li> |
| <li>Tasks should appear in the details of a framework</li> |
| <li>Check the stdout and stderr of the sandbox of failed tasks</li> |
| </ul> |
| </li> |
| <li>Mesos logs |
| <ul> |
| <li>Master and slave logs are both in <code>/var/log/mesos</code> by default</li> |
| </ul> |
| </li> |
| </ul> |
| |
| <p>And common pitfalls:</p> |
| |
| <ul> |
| <li>Spark assembly not reachable/accessible |
| <ul> |
| <li>Slaves must be able to download the Spark binary package from the <code>http://</code>, <code>hdfs://</code> or <code>s3n://</code> URL you gave</li> |
| </ul> |
| </li> |
| <li>Firewall blocking communications |
| <ul> |
| <li>Check for messages about failed connections</li> |
| <li>Temporarily disable firewalls for debugging and then poke appropriate holes</li> |
| </ul> |
| </li> |
| </ul> |
| |
| |
| </div> |
| |
| <!-- /container --> |
| </div> |
| |
| <script src="js/vendor/jquery-1.8.0.min.js"></script> |
| <script src="js/vendor/bootstrap.min.js"></script> |
| <script src="js/vendor/anchor.min.js"></script> |
| <script src="js/main.js"></script> |
| |
| <!-- MathJax Section --> |
| <script type="text/x-mathjax-config"> |
| MathJax.Hub.Config({ |
| TeX: { equationNumbers: { autoNumber: "AMS" } } |
| }); |
| </script> |
| <script> |
| // Note that we load MathJax this way to work with local file (file://), HTTP and HTTPS. |
| // We could use "//cdn.mathjax...", but that won't support "file://". |
| (function(d, script) { |
| script = d.createElement('script'); |
| script.type = 'text/javascript'; |
| script.async = true; |
| script.onload = function(){ |
| MathJax.Hub.Config({ |
| tex2jax: { |
| inlineMath: [ ["$", "$"], ["\\\\(","\\\\)"] ], |
| displayMath: [ ["$$","$$"], ["\\[", "\\]"] ], |
| processEscapes: true, |
| skipTags: ['script', 'noscript', 'style', 'textarea', 'pre'] |
| } |
| }); |
| }; |
| script.src = ('https:' == document.location.protocol ? 'https://' : 'http://') + |
| 'cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML'; |
| d.getElementsByTagName('head')[0].appendChild(script); |
| }(document)); |
| </script> |
| </body> |
| </html> |