Place more emphasis on using precompiled binary versions of Spark and Mesos instead of encouraging the reader to compile from source. Author: Andrew Ash <andrew@andrewash.com> Closes #756 from ash211/spark-1818 and squashes the following commits: 7ef3b33 [Andrew Ash] Brief explanation of the interactions between Spark and Mesos e7dea8e [Andrew Ash] Add troubleshooting and debugging section 956362d [Andrew Ash] Don't need to pass spark.executor.uri into the spark shell de3353b [Andrew Ash] Wrap to 100char 7ebf6ef [Andrew Ash] Polish on the section on Mesos Master URLs 3dcc2c1 [Andrew Ash] Use --tgz parameter of make-distribution 41b68ed [Andrew Ash] Period at end of sentence; formatting on :5050 8bf2c53 [Andrew Ash] Update site.MESOS_VERSIOn to match /pom.xml 74f2040 [Andrew Ash] SPARK-1818 Freshen Mesos documentation
8.6 KiB
layout | title |
---|---|
global | Running Spark on Mesos |
Why Mesos
Spark can run on hardware clusters managed by Apache Mesos.
The advantages of deploying Spark with Mesos include:
- dynamic partitioning between Spark and other frameworks
- scalable partitioning between multiple instances of Spark
How it works
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.
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.
To get started, follow the steps below to install Mesos and deploy Spark jobs via Mesos.
Installing Mesos
Spark {{site.SPARK_VERSION}} is designed for use with Mesos {{site.MESOS_VERSION}} and does not require any special patches of Mesos.
If you already have a Mesos cluster running, you can skip this Mesos installation step.
Otherwise, installing Mesos for Spark is no different than installing Mesos for use by other frameworks. You can install Mesos using either prebuilt packages or by compiling from source.
Prebuilt packages
The Apache Mesos project only publishes source package releases, no binary releases. But other third party projects publish binary releases that may be helpful in setting Mesos up.
One of those is Mesosphere. To install Mesos using the binary releases provided by Mesosphere:
- Download Mesos installation package from downloads page
- Follow their instructions for installation and configuration
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.
From source
To install Mesos directly from the upstream project rather than a third party, install from source.
- Download the Mesos distribution from a mirror
- Follow the Mesos Getting Started page for compiling and installing Mesos
Note: 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), you should also pass the
--prefix
option to configure
to tell it where to install. For example, pass
--prefix=/home/user/mesos
. By default the prefix is /usr/local
.
Verification
To verify that the Mesos cluster is ready for Spark, navigate to the Mesos master webui at port
:5050
Confirm that all expected machines are present in the slaves tab.
Connecting Spark to Mesos
To use Mesos from Spark, you need a Spark distribution available in a place accessible by Mesos, and a Spark driver program configured to connect to Mesos.
Uploading Spark Distribution
When Mesos runs a task on a Mesos slave for the first time, that slave must have a distribution of Spark available for running the Spark Mesos executor backend. A distribution of Spark is just a compiled binary version of Spark.
The Spark distribution can be hosted at any Hadoop URI, including HTTP via http://
, Amazon Simple
Storage Service via s3://
, or HDFS via hdfs:///
.
To use a precompiled distribution:
- Download a Spark distribution from the Spark download page
- Upload to hdfs/http/s3
To host on HDFS, use the Hadoop fs put command: hadoop fs -put spark-{{site.SPARK_VERSION}}.tar.gz /path/to/spark-{{site.SPARK_VERSION}}.tar.gz
Or if you are using a custom-compiled version of Spark, you will need to create a distribution using
the make-distribution.sh
script included in a Spark source tarball/checkout.
- Download and build Spark using the instructions here
- Create a Spark distribution using
make-distribution.sh --tgz
. - Upload archive to http/s3/hdfs
Using a Mesos Master URL
The Master URLs for Mesos are in the form mesos://host:5050
for a single-master Mesos
cluster, or zk://host:2181
for a multi-master Mesos cluster using ZooKeeper.
The driver also needs some configuration in spark-env.sh
to interact properly with Mesos:
- In
spark.env.sh
set some environment variables:
export MESOS_NATIVE_LIBRARY=<path to libmesos.so>
. This path is typically<prefix>/lib/libmesos.so
where the prefix is/usr/local
by default. See Mesos installation instructions above. On Mac OS X, the library is calledlibmesos.dylib
instead oflibmesos.so
.export SPARK_EXECUTOR_URI=<path to spark-{{site.SPARK_VERSION}}.tar.gz uploaded above>
.
- Also set
spark.executor.uri
to <path to spark-{{site.SPARK_VERSION}}.tar.gz>
Now when starting a Spark application against the cluster, pass a mesos://
or zk://
URL as the master when creating a SparkContext
. For example:
{% highlight scala %} val conf = new SparkConf() .setMaster("mesos://HOST:5050") .setAppName("My app") .set("spark.executor.uri", "<path to spark-{{site.SPARK_VERSION}}.tar.gz uploaded above>") val sc = new SparkContext(conf) {% endhighlight %}
When running a shell the spark.executor.uri
parameter is inherited from SPARK_EXECUTOR_URI
, so
it does not need to be redundantly passed in as a system property.
{% highlight bash %} ./bin/spark-shell --master mesos://host:5050 {% endhighlight %}
Mesos Run Modes
Spark can run over Mesos in two modes: "fine-grained" (default) and "coarse-grained".
In "fine-grained" mode (default), each Spark task runs as a separate Mesos task. This allows multiple instances of Spark (and other frameworks) to share machines at a very fine granularity, where each application gets more or fewer machines 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.
The "coarse-grained" mode will instead launch only one long-running Spark task on each Mesos machine, and dynamically schedule its own "mini-tasks" within it. The benefit is much lower startup overhead, but at the cost of reserving the Mesos resources for the complete duration of the application.
To run in coarse-grained mode, set the spark.mesos.coarse
property in your
SparkConf:
{% highlight scala %} conf.set("spark.mesos.coarse", "true") {% endhighlight %}
In addition, for coarse-grained mode, you can control the maximum number of resources Spark will
acquire. By default, it will acquire all cores in the cluster (that get offered by Mesos), which
only makes sense if you run just one application at a time. You can cap the maximum number of cores
using conf.set("spark.cores.max", "10")
(for example).
Running Alongside Hadoop
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 hdfs:// URL is required
(typically hdfs://<namenode>:9000/path
, but you can find the right URL on your Hadoop Namenode web
UI).
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 Hadoop on Mesos.
In either case, HDFS runs separately from Hadoop MapReduce, without being scheduled through Mesos.
Troubleshooting and Debugging
A few places to look during debugging:
- Mesos master on port
:5050
- Slaves should appear in the slaves tab
- Spark applications should appear in the frameworks tab
- Tasks should appear in the details of a framework
- Check the stdout and stderr of the sandbox of failed tasks
- Mesos logs
- Master and slave logs are both in
/var/log/mesos
by default
- Master and slave logs are both in
And common pitfalls:
- Spark assembly not reachable/accessible
- Slaves need to be able to download the distribution
- Firewall blocking communications
- Check for messages about failed connections
- Temporarily disable firewalls for debugging and then poke appropriate holes