hmda.init: Initialize or Restart H2O Cluster for HMDA Analysis

In HMDA: Holistic Multimodel Domain Analysis for Exploratory Machine Learning

Initialize or Restart H2O Cluster for HMDA Analysis

Description

Initializes or restarts an H2O cluster configured for Holistic Multimodel Domain Analysis. It sets up the cluster with specified CPU threads, memory, and connection settings. It first checks for an existing cluster, shuts it down if found, and then repeatedly attempts to establish a new connection, retrying up to 10 times if necessary.

Usage

hmda.init(
  cpu = -1,
  ram = NULL,
  java = NULL,
  ip = "localhost",
  port = 54321,
  verbatim = FALSE,
  restart = TRUE,
  shutdown = FALSE,
  ignore_config = TRUE,
  bind_to_localhost = FALSE,
  ...
)

Arguments

cpu	integer. The number of CPU threads to use. -1 indicates all available threads. Default is -1.
ram	numeric. Minimum memory (in GB) for the cluster. If NULL, all available memory is used.
java	character. Path to the Java JDK. If provided, sets JAVA_HOME accordingly.
ip	character. The IP address for the H2O server. Default is "localhost".
port	integer. The port for the H2O server. Default is 54321.
verbatim	logical. If TRUE, prints detailed cluster info. Default is FALSE.
restart	logical. if TRUE, the server is erased and restarted
shutdown	logical. if TRUE, the server is closed
ignore_config	logical. If TRUE, ignores any existing H2O configuration. Default is TRUE.
bind_to_localhost	logical. If TRUE, restricts access to the cluster to the local machine. Default is FALSE.
...	Additional arguments passed to h2o.init().

Details

The function sets JAVA_HOME if a Java path is provided. It checks for an existing cluster via h2o.clusterInfo(). If found, the cluster is shut down and the function waits 5 seconds. It then attempts to initialize a new cluster using h2o.init() with the specified settings. On failure, it retries every 3 seconds, up to 10 attempts. If all attempts fail, an error is thrown.

Value

An object representing the connection to the H2O cluster.

Author(s)

E. F. Haghish

Examples

## Not run: 
  # Example 1: Initialize the H2O cluster with default settings.
  library(hmda)
  hmda.init()

  # Example 2: Initialize with specific settings such as Java path.
  conn <- hmda.init(
      cpu = 4,
      ram = 8,
      java = "/path/to/java",     #e.g., "C:/Program Files/Java/jdk1.8.0_241"
      ip = "localhost",
      port = 54321,
      verbatim = TRUE
  )

  # check the status of the h2o connection
  h2o::h2o.clusterInfo(conn) #you can use h2o functions to interact with the server

## End(Not run)

HMDA documentation built on April 4, 2025, 6:06 a.m.