Skip to main content

Aerospike Interactive Notebooks

Aerospike is a distributed database designed to serve global applications with low latency, fast throughput, and resiliency to failures.

This section contains Jupyter notebooks for:

  • Spark: These notebooks show how Aerospike can be used in conjunction with Spark. Explore them staticly on this site or refer to the instructions below on how to set up and run Spark notebooks on Linux (CentOS) and MacOS X.

View the list of all notebooks here.

The download and use of this Aerospike software is governed by Aerospike Evaluation License Agreement.

Spark Notebooks#

Spark notebooks can run on Linux (CentOS) and MacOS X.

Set up Spark Notebooks on Linux (CentOS)#

yum installer used below - use dbpkg/rpm/other if your Linux distribution does not support yum

sudo yum -y install gcc zlib-devel openssl-devel libffi-devel sqlite-devel bzip2-devel bzip2 xz-devel screen wget

Get your own local copy of Python 3.7 (ignore if you have it already). Below we install to ~/.localpython

PYTHON_VERSION=3.7.1wget${PYTHON_VERSION}/Python-${PYTHON_VERSION}.tgztar zxvf Python-${PYTHON_VERSION}.tgzcd Python-${PYTHON_VERSION}mkdir ~/.localpython./configure --prefix=$HOME/.localpythonmakemake install

Set up a virtual Python environment - this is a sandbox which avoids you making system wide changes

# Install virtualenv tool~/.localpython/bin/pip3 install virtualenv# Create on-disk representation of virtual environment at ~/spark-venv~/.localpython/bin/virtualenv ~/spark-venv# Activate virtual environmentsource ~/spark-venv/bin/activate

Use of a virtual environment is indicated in the command line string - the name of the virtual environment - spark-env is added to the command line prompt - e.g.,

(spark-venv) [ec2-user@ip-10-0-0-248 Python-3.7.1]$

You can return to the system enviroment by typing deactivate and reactivate using source ~/spark-venv/bin/activate

Get rid of annoying messages concerning pip upgrade

pip install --upgrade pip

Note at this point, all our Python related tooling is local to our virtual environment. So which pip will give


Install required Python dependencies

pip install jupyter PySpark findspark numpy pandas matplotlib sklearn

If you plan on using scala in your workbooks you need to install the spylon kernel - some care is needed with Python versioning

pip install spylon_kernelPYTHON=$(which python)sudo $PYTHON -m spylon_kernel install

Install Spark and set $SPARK_HOME. Note you may need to change the SPARK_VERSION if you get a 404 following the wget.

SPARK_VERSION=2.4.7HADOOP_VERSION=2.7cd /tmpwget${SPARK_VERSION}/spark-${SPARK_VERSION}-bin-hadoop${HADOOP_VERSION}.tgztar xvfz spark-${SPARK_VERSION}-bin-hadoop${HADOOP_VERSION}.tgzsudo mv spark-${SPARK_VERSION}-bin-hadoop${HADOOP_VERSION} /opt/export SPARK_HOME=/opt/spark-${SPARK_VERSION}-bin-hadoop${HADOOP_VERSION}export PYTHONPATH=$SPARK_HOME/python:$PYTHONPATHcd ~

Use of the Aerospike Spark Connector requires a valid feature key. The notebooks assume this is located at /etc/aerospike/features.conf. Make sure your feature key is locally available, and if it is not located as above, modify the AS_FEATURE_KEY_PATH variable at the head of the notebook. You may need to run

sudo mkdir /etc/aerospikesudo chmod 777 /etc/aerospike

Make sure you have the interactive-notebooks repository locally.

git clone

Finally start Jupyter. Change the IP in the string below - it can be localhost, but if you want to access from a remote host, choose the IP of one of your ethernet interfaces. You could replace with $(hostname -I | awk '{print $1}')

Note I set the notebook-dir to point to the directory containing the notebooks in this repository. You also will need SPARK_HOME and PYTHONPATH set correctly (reproducing the former from the above).

SPARK_VERSION=2.4.7HADOOP_VERSION=2.7export SPARK_HOME=/opt/spark-${SPARK_VERSION}-bin-hadoop${HADOOP_VERSION}export PYTHONPATH=$SPARK_HOME/python:$PYTHONPATHjupyter notebook --no-browser --ip=<IP> --port=8888 --notebook-dir=~/interactive-notebooks/spark/

You will see output similar to

[I 09:36:52.202 NotebookApp] Writing notebook server cookie secret to /home/ec2-user/.local/share/jupyter/runtime/notebook_cookie_secret[I 09:36:52.370 NotebookApp] Serving notebooks from local directory: /home/ec2-user/interactive-notebooks/spark[I 09:36:52.370 NotebookApp] Jupyter Notebook 6.1.4 is running at:[I 09:36:52.370 NotebookApp][I 09:36:52.370 NotebookApp]  or[I 09:36:52.371 NotebookApp] Use Control-C to stop this server and shut down all kernels (twice to skip confirmation).[C 09:36:52.373 NotebookApp]

You will need to use the URLs in the output to access jupyter - as the security token is expected.

You can omit this step by omitting the --no-browser flag - in that case jupyter will open a browser window local to itself, and request the Notebook app URL above.

You may wish to run the jupyter startup command from a screen so it will stay running if your session terminates. We installed screen at the outset to allow for this.

pyenv / Linux#

You can go down the pyenv route on Linux as per the instructions for Mac. You install pyenv differently

sudo yum -y install gcc git zlib-devel openssl-devel libffi-devel sqlite-devel bzip2-devel bzip2 xz-devel screengit clone .pyenvexport PATH=$PATH:~/.pyenv/bin

but once done, just pick up the MacOS instructions at pyenv install 3.7.3

Set Up Spark Notebooks on MacOS X#

The main challenge is getting a sufficiently up to date version of Python installed and set as your working version. You mustn't mess with your existing version of Python (see xkcd).

pyenv is the tool to help with this.

First you'll need brew the package manager for macOS. From instructions

/bin/bash -c "$(curl -fsSL"

Next install pyenv

brew install pyenv

and finally we can install our required python version. The subsequent 'global' command sets 3.7.3 as our selected version

pyenv install 3.7.3pyenv global 3.7.3

The command below sets up our path so the required version of Python is used. Once done, do python --version to check.

eval "$(pyenv init -)"

You can now set up your virtual environment - this is a sandbox which avoids you making system wide changes. Note this is the same as the steps above for Linux, except we don't have to give explicit paths to pip, virtualenv.

# Install virtualenv toolpip install virtualenv# Create on-disk representation of virtual environment at ~/spark-venvvirtualenv ~/spark-venv# Activate virtual environmentsource ~/spark-venv/bin/activate

You can now follow the Linux instructions from

pip install jupyter PySpark findspark numpy pandas matplotlib sklearn