Implement Node and Pod Affinity/Anti-Affinity in Kubernetes: A Practical Example

I introduced the concept of node and pod affinity/anti-affinity in last week’s tutorial. We will explore the idea further through a real-world scenario.

Objective

We are going to deploy three microservices — MySQL, Redis, and a Python/Flask web app in a four-node Kubernetes cluster. Since one of the nodes is attached to SSD disk, we want to ensure that the MySQL Pod is scheduled on the same node. Redis is used to cache the database queries to accelerate application performance. But no node will run more than one Pod of Redis. Since Redis is utilized as a cache, it doesn’t make sense to run more than one Pod per node. The next goal is to make sure that the web Pod is placed on the same node as the Redis Pod. This will ensure low latency between the web and the cache layer. Even if we scale the number of replicas of the web Pod, it will never get placed on a node that doesn’t have Redis Pod.

Read the entire article at The New Stack

Janakiram MSV is an analyst, advisor, and architect. Follow him on Twitter,  Facebook and LinkedIn.