Using Spark 2 from R

R users can access Spark 2 using sparklyr. Although Cloudera does not ship or support sparklyr, we do recommend using sparklyr as the R interface for Cloudera Data Science Workbench.

Installing sparklyr

Install the latest version of sparklyr from GitHub as follows. The latest GitHub package includes a patch that ensures compatibility with Cloudera's Distribution of Apache Spark 2. In contrast, an installation from CRAN will not include this capability.

devtools::install_github("rstudio/sparklyr")

Connecting to Spark 2

You can connect to local instances of Spark 2 as well as remote clusters.

## Connecting to Spark 2
# Connect to an existing Spark 2 cluster in YARN client mode using the spark_connect
# function.
library(sparklyr)
system.time(sc <- spark_connect(master = "yarn-client"))
# The returned Spark 2 connection (sc) provides a remote dplyr data source to the Spark 2 cluster.