Should I use NFS for Persistent Volumes when running RabbitMQ on a Kubernetes cluster?

Question

Amazon Web Services

Hello,

is there such a task - to raise a RabbitMQ cluster in kubernetes, so that messages are not lost when one Availability Zone in AWS falls? I'm thinking of raising EFS and connecting Persistent Storage from Kubernetes there using EFS Provisioner ... is it worth doing this and will it help to create a cluster?

Your opinion !

all the best,
Vadim

Answer 1

This is a very stupid idea for the following reasons:

• EFS is network storage and is slower than EBS
  
• EFS is needed to share data between multiple network resources when creating a cluster involves isolated instances
  
• Messages will not be lost if they are already saved on EBS, but you should not forget about backups
  
• Using RabbitMQ on AWS is generally expensive and useless when you have SQS, SNS, EventBridge, Kinesis, MSK, Amazon MQ, ...
  

Answer 2

I don't agree with your answer!
EFS is needed to share data between multiple network resources when creating a cluster involves isolated instances - this is exactly what is required, because everything works in Kubernetes and RabbitMQ pods will be in different availability zones.
Messages will not be lost if they are already saved on EBS, but you should not forget about backups ... well, if the zone falls, what will we do? How well do you know Rabbit - when storage is lost, messages are lost, or if there is mirroring, will it still continue to work?
Using RabbitMQ on AWS is generally expensive and useless when you have SQS, SNS, EventBridge, Kinesis, MSK, Amazon MQ, ... - I agree, but no one will rewrite the code, a lot of legacy has accumulated

Answer 3

I am in favor of using Mirrored Queue and other built-in things for HA ( https://www.rabbitmq.com/ha.html ).