How to install H2O R package on an offline server?

Question

I had to install H2O R package on an offline server. I created a local repo via miniCRAN and tried to install the package from it but installation failed with this error:

Error in download.file(md5_url, destfile = md5_file, mode = "w", cacheOK = FALSE,  :
  cannot open URL 'http://s3.amazonaws.com/h2o-release/h2o/rel-vajda/3/Rjar/h2o.jar.md5'
Error : unable to load R code in package ‘h2o’
ERROR: lazy loading failed for package ‘h2o’
* removing ‘/usr/lib64/R/library/h2o’

I circumvented this by manual transferring of installed binaries from an online server but this is obviously just a workaround. So is there a normal way to install H2O R package on an offline server?

Answer 1

The R install package is embedded in the zip download file.

This will work, assuming all the dependencies have also been downloaded and installed.

(Of course, substitute the specific version that you want.)

wget https://h2o-release.s3.amazonaws.com/h2o/rel-weierstrass/2/h2o-3.14.0.2.zip
unzip h2o-3.14.0.2.zip 
cd h2o-3.14.0.2
cd R
R CMD INSTALL h2o_3.14.0.2.tar.gz 

Answer 2

You have to have h2o.jar already downloaded.

Then set environment variable like so:

Sys.setenv(H2O_JAR_PATH="_your_h2o_installation_path_/h2o.jar").

Answer 3

This worked for me.

First download the h2o package archive and get the Jar file then set up H2O_JAR_PATH using this command

Sys.setenv(H2O_JAR_PATH="/home/hadoop/R/h2o.jar")

Finally,

install.packages("~/R/h2o_3.20.0.8.tar.gz", repos = NULL, type = "source")
library(h2o)