SpecterOps-Nemesis/docs/rabbitmq.md

Design decisions

• Didn't go with the RabbitMQ operator because it doesn't play nicely with skaffold and is a fairly simple wrapper for the statefulset
  • Doesn't delete pods/statefulsets during shutdown
• Didn't go with just the docker container because to scale up you need to deploy them one at a time (so used a statefulset).

Scaling

Rabbitmq scales vertifcally better (so give it more CPU + RAM) Add more replicas by editing ./helm/nemesis/templates/rabbitmq.service.yaml.

Performance testing

kubectl run perf-test -it --rm  --image=pivotalrabbitmq/perf-test -- --uri amqp://rabbit:Qwerty12345@nemesis-rabbitmq-svc

Plumber example usage:

plumber read rabbit --address="amqp://nemesis:Qwerty12345@192.168.230.42:5672" --queue-name="file_data" --exchange-name="nemesis" --binding-key=" " --queue-durable -f

References

• https://blog.rabbitmq.com/posts/2020/08/deploying-rabbitmq-to-kubernetes-whats-involved/
• https://github.com/GoogleCloudPlatform/click-to-deploy/tree/master/k8s/rabbitmq