Skip to content

Latest commit

 

History

History
95 lines (68 loc) · 4.37 KB

13_spark_shell.md

File metadata and controls

95 lines (68 loc) · 4.37 KB

Documentation

Using the Spark Cassandra Connector with the Spark Shell

These instructions were last confirmed with Cassandra 3.0.9, Spark 2.0.2 and Connector 2.0.1.

For this guide, we assume an existing Cassandra deployment, running either locally or on a cluster, a local installation of Spark and an optional Spark cluster. For detail setup instructions see setup spark-shell

To use the Spark Cassandra Connector from within the Spark Shell, we need to load the Connector and all its dependencies in the shell context. The easiest way to achieve that is to build an assembly (also known as "fat jar") that packages all dependencies.

Starting the Spark Shell

If you don't include the master address below the spark shell will run in Local mode. For the package be sure to pick the version of Scala that your Spark build uses (The "2.1X" portion of the package. If you aren't sure for Spark < 2.0 use 2.10).

Find additional versions at Spark Packages

cd spark/install/dir
#Include the --master if you want to run against a spark cluster and not local mode
./bin/spark-shell [--master sparkMasterAddress] --jars yourAssemblyJar --packages datastax:spark-cassandra-connector:2.0.1-s_2.11 --conf spark.cassandra.connection.host=yourCassandraClusterIp

By default spark will log everything to the console and this may be a bit of an overload. To change this copy and modify the log4j.properties template file

cp conf/log4j.properties.template conf/log4j.properties

Changing the root logger at the top from INFO to WARN will significantly reduce the verbosity.

Example

Import connector classes

import com.datastax.spark.connector._ //Imports basic rdd functions
import com.datastax.spark.connector.cql._ //(Optional) Imports java driver helper functions

Test it out

val c = CassandraConnector(sc.getConf)
c.withSessionDo ( session => session.execute("CREATE KEYSPACE test WITH replication={'class':'SimpleStrategy', 'replication_factor':1}"))
c.withSessionDo ( session => session.execute("CREATE TABLE test.fun (k int PRIMARY KEY, v int)"))


// Your results may differ 
//res1: Array[com.datastax.spark.connector.CassandraRow] = Array(CassandraRow{k: 60, v: 60}, CassandraRow{k: 67, v: 67}, CassandraRow{k: 10, v: 10})

Creating a playground with Docker

We can use Docker to quickly create a working setup without the need to install and configure Cassandra.

For this guide we will use the Cassandra docker image maintained by Al Tobert on GitHub

First, let's pull the docker image:

docker pull tobert/cassandra

To instantiate a Cassandra container, we need to specify a host volume where the data will be stored. We store the resulting container id in the variable CASSANDRA_CONTAINER_ID. We will need that id afterwards to lookup the IP address of the running instance.

mkdir /srv/cassandra
CASSANDRA_CONTAINER_ID=`docker run -d -v /srv/cassandra:/data tobert/cassandra`

Alternatively, if the data is disposable (e.g. tests), we can omit the volume, resulting in the following step:

CASSANDRA_CONTAINER_ID=$(docker run -d tobert/cassandra)

Now, we obtain the IP Address that docker assigned to the container:

CASSANDRA_CONTAINER_IP=$(docker inspect -f '{{ .NetworkSettings.IPAddress }}' $CASSANDRA_CONTAINER_ID)

We could also run a named container using --name and query for IP using the name, e.g.

$ docker run --name cassie -d tobert/cassandra
$ docker ps
CONTAINER ID        IMAGE               COMMAND                  CREATED             STATUS              PORTS                                               NAMES
3e9a39418f6a        tobert/cassandra    "/bin/cassandra-docke"   2 seconds ago       Up 2 seconds        7000/tcp, 7199/tcp, 9042/tcp, 9160/tcp, 61621/tcp   cassie
$ CASSANDRA_CONTAINER_IP=$(docker inspect -f '{{ .NetworkSettings.IPAddress }}' cassie)

And we can start the spark shell using that running container:

cd spark/install/dir
#Include the --master if you want to run against a spark cluster and not local mode
./bin/spark-shell [--master sparkMasterAddress] --jars yourAssemblyJar --conf spark.cassandra.connection.host=$CASSANDRA_CONTAINER_IP

Next - DataFrames