Home

Cluster wide report

This topic explains how to set up and run Unravel's Cluster Wide Report.

Cluster Wide Report is a java App designed to help you fine-tune your cluster wide parameters to maximize its efficiency based upon your cluster's typical workload.

Cluster Wide Report

  • Collects performance data of prior completed jobs.

  • Analyzes the jobs relative to the cluster's current configuration.

  • Generates recommended cluster parameter changes.

  • Predicts and quantifies the impact the changes will have on future runs of the jobs.

Most of the recommendations revolve around the parameters

  • MapSplitSizeParams

  • HiveExecReducersBytesParam

  • HiveExecParallelParam

  • MapReduceSlowStartParam

  • MapReduceMemoryParams

You can choose to implement some or all of the recommended settings.

Step-by-step guide

  1. Download the report tarball from the Unravel preview site.

    curl -v https://preview.unraveldata.com/img/ClusterReportSetup.tar.gz -o ClusterReportSetup.tar.gz

  2. Unpack and run the setup script to install the app in /usr/local/unravel/install_bin.

    tar zxvf ClusterReportSetup.tar.gz
cd ClusterReportSetup
sudo ./setup.sh /usr/local/unravel/install_bin

  3. The app is now installed in /usr/local/unravel/install_bin/ClusterReport. cd to the installation directory.

    ls
dbin/             etc/             origJars/               
dlib/             logs/            origJars.tar.gz

  4. cd to dbin and edit Input.txt.

    cd /usr/local/unravel/install_bin/ClusterReport/dbin 
vi Input.txt

  5. Configure Input.txt for your cluster and report parameters.

    cluster_id =
queue =
start_date=2018-01-01
end_date=2018-03-28
mapreduce.map.memory.mb=2048
mapreduce.reduce.memory.mb=2048
hive.exec.reducers.bytes.per.reducer=268435456
mapreduce.input.fileinputformat.split.maxsize=256000000

  6. Run the report.

    su - hdfs ./cluster_report.sh
In this section: