Redis k8s operator deployment on Kind

By default, this set of scripts will set up a fully functioning Redis Active-Active database distributed over 2 Redis clusters running on seperate vanilla/community k8s clusters deployed in kind (Kubernetes in Docker).

Requirements

A single Linux - Debian 12 or Ubuntu 24 - x86 host with enough resources (see at the bottom for validated configurations). Some tools and OS tuning is required. Run the prereqs/install-fresh script and Reboot.

Getting started quickly

• Run the create-all.sh script.
• Wait 5-10 mins depending on the hardware.

Stopping quickly

Run the delete-all.sh script.

What it deploys

Basics:

• A configurable number (default = 2) k8s clusters in Kind
• Deploy Redis Enterprise (operator) in all clusters.
• Create Redis clusters in all k8s clusters.
• Create a Redis database (geo-replicated CRDB in case of Active/Active).

Optional:

• Deploy the Redis Enterprise admission controller.
• Rack zone aware deployment.
• a loadbalancer (metallb) to let the clusters talk to each other through a "public" endpoint (in the Docker kind network).
• an ingress controller (ingress-nginx / haproxy-ingress) to facilitate Redis Operator created Ingress resources.
• patch k8s DNS (coredns) to resolve the (remote) cluster api and database.
• monitoring through Prometheus + Grafana and preconfigured Redis dashboards.

Good to know

• There are multiple haproxy ingresses. One from HAProxy Inc. and one from jcmoraisjr. This is tested with the latter one.
• There are multiple nginx ingresses. So far this is tested with the kubernetes 'ingress-nginx'.
• There are more Ingresses that work, such as the Nginx Inc. (F5) Ingress Nginx Controller and Contour. However, they require (manual) creation of CRDs for SSL/TLS passthrough if regular Ingress Kubernetes resources are not supporting passthrough. E.g. TransportServer for Nginx Inc. or HTTPProxy for Contour.

Caveats / todos

• DNS patching is hardwired to api and a named 'db1' database.
• This is a scripted deployment, so not all steps are idempotent. If you change the config, you may need to tear down and recreate to get consistent again.

Minimum hardware requirements

The minimum proper setup with 2 participating Redis clusters (3 k8s worker nodes each) requires 16GB of RAM, a 30GB OS disk, and >4 physical CPU cores.

Some tested configs that work:

• physical i5-9600K (6 cores @ 4.5 GHz)
• GCP n2-standard-8 (8 vCPU - 4 cores @ 2.1-2.8 GHz)
• AWS t3.2xlarge (8 vCPU @ 2.5 GHz)

What won't work:

• GCP n2-standard-4 (4 vCPU - 2 cores @ 2.1-2.8 GHz)
• GCP c2-standard-4 (4 vCPU - 2 cores @ 3.1 GHz)
• AWS t3.xlarge (4 vCPU @ 2.5 GHz)