This section describes how Spark connects to Cassandra and how to execute CQL statements from Spark applications.
To connect your Spark application to Cassandra, set connection options in the
SparkConf object. These are prefixed with spark. so that they can be recognized
from the spark-shell and set within the $SPARK_HOME/conf/spark-default.conf.
Example:
val conf = new SparkConf(true)
.set("spark.cassandra.connection.host", "192.168.123.10")
.set("spark.cassandra.auth.username", "cassandra")
.set("spark.cassandra.auth.password", "cassandra")
val sc = new SparkContext("spark://192.168.123.10:7077", "test", conf)See the reference secition for a full list of options Cassandra Connection Parameters
To import Cassandra-specific functions on SparkContext and RDD objects, call:
import com.datastax.spark.connector._ Query retry delay can be configured in few different ways:
<delay in ms>- for a constant delay before each retry<initial delay in ms>+<increase in ms>- for a linearly increasing delay - delay before each retry will be longer than the delay before the previous retry by increase factor<initial delay in ms>*<increase multiplier, float>- for an exponentially increasing delay - delay before each retry will be as many times longer than the previous retry delay as specified by the multiplier
Whenever you call a method requiring access to Cassandra, the options in the SparkConf object will be used
to create a new connection or to borrow one already open from the global connection cache.
The initial contact node given in
spark.cassandra.connection.host can be any node of the cluster. The driver will fetch the cluster topology from
the contact node and will always try to connect to the closest node in the same data center. If possible,
connections are established to the same node the task is running on. Consequently, good locality of data can be achieved and the amount
of data sent across the network is minimized.
Connections are never made to data centers other than the data center of spark.cassandra.connection.host.
If some nodes in the local data center are down and a read or write operation fails, the operation won't be retried on nodes in
a different data center. This technique guarantees proper workload isolation so that a huge analytics job won't disturb
the realtime part of the system.
Connections are cached internally. If you call two methods needing access to the same Cassandra cluster
quickly, one after another, or in parallel from different threads, they will share the same logical connection
represented by the underlying Java Driver Cluster object.
Eventually, when all the tasks needing Cassandra connectivity terminate,
the connection to the Cassandra cluster will be closed shortly thereafter. The period of time for keeping unused connections
open is controlled by the global spark.cassandra.connection.keep_alive_ms system property, which defaults to 250 ms.
If you ever need to manually connect to Cassandra in order to issue some CQL statements,
this driver offers a handy CassandraConnector class which can be initialized from the SparkConf object
and provides access to the Cluster and Session objects. CassandraConnector instances are serializable
and therefore can be safely used in lambdas passed to Spark transformations.
Assuming an appropriately configured SparkConf object is stored in the conf variable, the following
code creates a keyspace and a table:
import com.datastax.spark.connector.cql.CassandraConnector
CassandraConnector(conf).withSessionDo { session =>
session.execute("CREATE KEYSPACE test2 WITH REPLICATION = {'class': 'SimpleStrategy', 'replication_factor': 1 }")
session.execute("CREATE TABLE test2.words (word text PRIMARY KEY, count int)")
}