DAT500 MSc course

cd into setup

• Ensure that setup_env.sh has correct addresses for /etc/hosts/.
• Run sudo setup.sh, this will setup the environment, install necessary libraries and Hadoop.
• Copy files & folders in /scripts/* to /usr/local/hadoop/
• Ensure that Python 3 is the default Python interpreter! And that it is a version prior to 3.8.0.
• Ensure that Java 8 is the default Java version. Check ~/.bashrc and ~/.profile for potential Java 11 overrides.

• Run sudo setup_spark.sh, this will install Spark.
• Copy files & folders in /master-only/ to /usr/local/hadoop/
• Copy files & folders in /spark/ to /usr/local/spark/
• Update /usr/local/hadoop/etc/hadoop/workers to include your workers. (slave nodes)

• Copy files & folders in /slave-only/ to /usr/local/hadoop/

Format the namenode like this, hdfs namenode -format (type in terminal)

• Start Hadoop and Spark, run start-dfs.sh, start-yarn.sh, and start-history-server.sh.
• Stop Hadoop and Spark, run stop-history-server.sh, stop-yarn.sh and stop-dfs.sh.

To test if Hadoop is working, simply run setup/setup_runtest.sh.

• Regular MapReduce, /run/hadoop-streaming.sh <file_mapper.py> <file_reducer.py>
• With MRJob, python3 some_file.py --hadoop-streaming-jar /usr/local/hadoop/share/hadoop/tools/lib/hadoop-streaming-3.1.1.jar -r hadoop hdfs:///data/input_data.csv --output-dir hdfs:///output/xyz --no-output
• With Spark, spark-submit --master yarn some_file.py

cd into src and execute run.sh (~/dat500/src/)

• Sync results retrieved from Hadoop & Spark
• cd into src
• conda install geopandas
• conda install geoplot -c conda-forge
• conda install -c conda-forge cartopy
• pip install -r requirements.txt
• Open visualize.ipynb to visualize results

(the CSV dataset was generated from the original census dataset by running src/merge_data.ipynb)

• Check if /usr/local/hadoop/etc/hadoop/hadoop-env.sh has any faulty paths.
• Python 3.6.9 was built from source, and should be the default Python version. However if a different version is preferred, update /usr/local/spark/conf/spark-env.sh to use python3.X for its Python drivers.

• Check https://codethief.io/hadoop101/ for similar cluster configuration / setup.