The latest live version is 3.1.0.17.
Previous Live Version: 3.0.3.5
To install mHub for Spark to your home directory simply download the tar and extract:
cd ~
wget https://syniti2.hosted-by-files.com/ddl/mHub_Spark_US_setup_3.1.0.17_linux.tar.gz
To extract:
tar -zxvf mHub_Spark_US_setup_3.1.0.17_linux.tar.gz
The tar will extract to a folder called mHubSpark. This folder contains the following:
| lib | Folder containing mHUB native shared object library. |
| libstdc++ | Folder containing C++ standard libraries required by mHUB. |
| samples | Folder containing sample applications. |
| HubSpark.jar | Jar containing the 360Science for Spark package. |
| matchITHub.jar | Jar containing the mHUB Java interface. |
| readme | Text file containing overview of product. |
| release_notes | Text file containing release notes. |
The native library files should be placed in folder, e.g. /usr/local/lib64, that can be synced with all the processing nodes. Execute the following steps:
cd mHubSpark/libchmod 777 ./libmatchITHub.sosudo cp libmatchITHub.so /usr/local/lib64cd ../libstdc++chmod 777 ./lib*.sosudo cp lib*.so /usr/local/lib64
Ensure that the files in /usr/local/lib64 are synced with all processing nodes in the cluster and that spark-env.sh contains these lines:
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/lib64SPARK_YARN_USER_ENV="LD_LIBRARY_PATH=$LD_LIBRARY_PATH"