Running Spark on Mesos
Spark can run on clusters managed by Apache Mesos. Follow the steps below to install Mesos and Spark:
- Download and build Spark using the instructions here. Note: Don’t forget to consider what version of HDFS you might want to use!
- Download, build, install, and start Mesos 0.13.0 on your cluster. You can download the Mesos distribution from a mirror. See the Mesos Getting Started page for more information. Note: If you want to run Mesos without installing it into the default paths on your system (e.g., if you don’t have administrative privileges to install it), you should also pass the
configureto tell it where to install. For example, pass
--prefix=/home/user/mesos. By default the prefix is
- Create a Spark “distribution” using
- Rename the
distdirectory created from
- Create a
tar czf spark-0.9.0-incubating.tar.gz spark-0.9.0-incubating
- Upload this archive to HDFS or another place accessible from Mesos via
http://, e.g., Amazon Simple Storage Service:
hadoop fs -put spark-0.9.0-incubating.tar.gz /path/to/spark-0.9.0-incubating.tar.gz
- Create a file called
confdirectory, by copying
conf/spark-env.sh.template, and add the following lines to it:
export MESOS_NATIVE_LIBRARY=<path to libmesos.so>. This path is usually
<prefix>/lib/libmesos.so(where the prefix is
/usr/localby default, see above). Also, on Mac OS X, the library is called
export SPARK_EXECUTOR_URI=<path to spark-0.9.0-incubating.tar.gz uploaded above>.
export MASTER=mesos://HOST:PORTwhere HOST:PORT is the host and port (default: 5050) of your Mesos master (or
zk://...if using Mesos with ZooKeeper).
- To run a Spark application against the cluster, when you create your
SparkContext, pass the string
mesos://HOST:PORTas the master URL. In addition, you’ll need to set the
spark.executor.uriproperty. For example:
val conf = new SparkConf() .setMaster("mesos://HOST:5050") .setAppName("My app") .set("spark.executor.uri", "<path to spark-0.9.0-incubating.tar.gz uploaded above>") val sc = new SparkContext(conf)
Mesos Run Modes
Spark can run over Mesos in two modes: “fine-grained” and “coarse-grained”. In fine-grained mode, which is the 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, but it comes with an additional overhead in launching each task, which may be inappropriate for low-latency applications (e.g. 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:
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, just use a hdfs:// URL (typically
hdfs://<namenode>:9000/path, but you can find the right URL on your Hadoop Namenode’s web UI).
In addition, it is possible to also run Hadoop MapReduce on Mesos, to get 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 going through Mesos.