Cassandra Playroom Overview

This is a simple setup of a Cassandra cluster with Reaper, Prometheus, Grafana as well as Neo4j running in Docker. The purpose of this project is to provide a "playroom" for those who want to familiarize themselves with Apache Cassandra and Neo4j.

Noteworthy

Neo4j enterprise is used with the "evaluation" license agreement which allows you 30 days to evaluate it. Set the NEO4J_ACCEPT_LICENSE_AGREEMENT environment variable in the docker-compose.yml to yes (from the default eval) if you whish to do so. Read more here about the different Neo4j editions and licensing stuff.

This project also has a jupyter notebook (cassandra.ipynb) that has some examples of how to interact with cassandra and neo4j.

You'll need docker preinstalled and you'll need at least 6GB of ram that can be dedicated to the docker containers - if one of the cassandra nodes unexpectedly dies, then it is probably because the docker daemon OOM killed it, it should automatically be restarted by the docker daemon (specified in docker-compose.yml). You can adjust the amount of ram used by cassandra changing MAX_HEAP_SIZE and NEW_HEAPSIZE in the docker-compose.yml file. You can also tweak the amount of ram allocated to each docker container by adjusting the mem_limit for the respective service entries in the docker-compose.yml files.

Metrics are collected through the JMX Exporter form Prometheus java agent (injected via a JAR on each cassandra node).

Config files for for the different components can be found in the config directory.

References

• Cassandra
• Neo4j
• Reaper
• JMX Exporter (for Prometheus)
• Prometheus
• Grafana

Run cluster

docker compose build
docker compose up -d

Check node statuses

NOTE: The cluster may take a few minutes to start up, depending on the machine running the containers.

docker exec -it tools nodetool -h cassandra1 status
docker exec -it tools nodetool -h cassandra2 status
docker exec -it tools nodetool -h cassandra3 status

Neo4j

Login to the Neo4j Browser with username neo4j and password password.

Reaper

Login to the Reaper UI with username admin and password admin.

Click on "Add Cluster" and fill in the following information:

• Seed node: cassandra1

Prometheus

You can connect to prometheus using this link

Grafana

Login to the Grafana UI with username admin and password admin.

Setup for python notebook

python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
pip install -r requirements.txt

Running the stress tool

Even though it does not make a lot of sense to stress test cassandra in this "playroom" context, below you can find an example of how to run the cassandra-stress tool.

# Writes only
docker exec -ti tools cassandra-stress user profile=/profiles/simple-stress.yaml ops\(insert=1\) -node cassandra1,cassandra2,cassandra3 -rate threads=8 -graph file=/reports/stress.html title=PlayroomStress revision=insert_run1

# Reads only
docker exec -ti tools cassandra-stress user profile=/profiles/simple-stress.yaml ops\(simple1=1\) -node cassandra1,cassandra2,cassandra3 -rate threads=8 -graph file=/reports/stress.html title=PlayroomStress revision=read_run1

# 10% Write and 90% Read
docker exec -ti tools cassandra-stress user profile=/profiles/simple-stress.yaml ops\(insert=1,simple1=9\) -node cassandra1,cassandra2,cassandra3 -rate threads=8 -graph file=/reports/stress.html title=PlayroomStress revision=rw_run1

# Copy the generated graph & report
docker cp tools:/reports/stress.html stress.html