Using SSL to secure H2O Flow

Sparkling Water supports security of H2O Flow user interface. There are two ways how to secure the Flow.

  • Provide the existing Java key store and password.

  • Let Sparkling Water automatically create the necessary files. This solution has several limitations which are described bellow.

Using existing Java keystore

In order to use https correctly, the following two options need to be specified:

  • spark.ext.h2o.jks - path to Java keystore file

  • spark.ext.h2o.jks.pass - keystore file password

Scala

To enable https in Sparkling Water, you can start Sparkling Water as:

bin/sparkling-shell --conf "spark.ext.h2o.jks=/path/to/keystore" --conf "spark.ext.h2o.jks.pass=password"

and when you have the shell running, start H2OContext as:

import org.apache.spark.h2o._
val hc = H2OContext.getOrCreate(spark)

You can also start Sparkling shell without the configuration and specify it using the setters on H2OConf as:

import org.apache.spark.h2o._
val conf = new H2OConf(spark).setJks("/path/to/keystore").setJksPass("password")
val hc = H2OContext.getOrCreate(spark, conf)

Python

To enable https in PySparkling, you can start PySparkling as:

bin/pysparkling --conf "spark.ext.h2o.jks=/path/to/keystore" --conf "spark.ext.h2o.jks.pass=password"

and when you have the shell running, start H2OContext as:

from pysparkling import *
hc = H2OContext.getOrCreate(spark)

You can also start PySparkling shell without the configuration and specify it using the setters on H2OConf as:

from pysparkling import *
conf = H2OConf(spark).setJks("/path/to/keystore").setJksPass("password)
hc = H2OContext.getOrCreate(spark, conf)

R

To enable https in RSparkling, run in RStudio:

library(rsparkling)
sc <- spark_connect(master = "local")
conf <- H2OConf(sc)$setJks("/path/to/keystore")$setJksPass("password")
hc <- H2OContext.getOrCreate(sc, conf)

In case your certificates are self-signed, the connection to the H2O cluster will fail due to the security limitations. In this case, you can skip the certificates verification by calling setVerifySslCertificates on H2OConf as:

Scala

val conf = new H2OConf(spark).setVerifySslCertificates(false)
val hc = H2OContext.getOrCreate(spark, conf)

Python

conf = H2OConf(spark).setVerifySslCertificates(False)
hc = H2OContext.getOrCreate(spark, conf)

R

conf <- H2OConf(sc)$setVerifySslCertificates(FALSE)
hc <- H2OContext.getOrCreate(sc, conf)

Generate the files automatically

Sparkling Water can generate the necessary key store and password automatically. To enable the automatic generation, the spark.ext.h2o.auto.flow.ssl option needs to be set to true. In this mode only self-signed certificates are created.

Scala

To enable the security using this mode in Sparkling Water, start Sparkling Shell as:

bin/sparkling-shell --conf "spark.ext.h2o.auto.flow.ssl=true" --conf "spark.ext.h2o.verify_ssl_certificates=false"

and when you have the shell running, start H2OContext as:

import org.apache.spark.h2o._
val hc = H2OContext.getOrCreate(spark)

You can also start Sparkling shell without the configuration and specify it using the setters on H2OConf as:

import org.apache.spark.h2o._
val conf = new H2OConf(spark).setAutoFlowSslEnabled().setVerifySslCertificates(false)
val hc = H2OContext.getOrCreate(spark, conf)

Python

To enable https in PySparkling using this mode, you can start PySparkling as:

bin/pysparkling --conf "spark.ext.h2o.auto.flow.ssl=true"  --conf "spark.ext.h2o.verify_ssl_certificates=false"

and when you have the shell running, start H2OContext as:

from pysparkling import *
hc = H2OContext.getOrCreate(spark)

You can also start PySparkling shell without the configuration and specify it using the setters on H2OConf as:

from pysparkling import *
conf = H2OConf(spark).setAutoFlowSslEnabled().setVerifySslCertificates(False)
hc = H2OContext.getOrCreate(spark, conf)

R

To enable https in RSparkling using this mode, run in your RStudio:

library(rsparkling)
sc <- spark_connect(master = "local")
conf <- H2OConf(sc)$setAutoFlowSslEnabled()$setVerifySslCertificates(FALSE)
hc <- H2OContext.getOrCreate(sc, conf)