I've successfully completed a proof of concept for integrating Kong into our system as an API gateway. I've experimented with hybrid mode using a PostgreSQL database and the Kong Helm chart. Now, I'm gearing up for the production deployment. What should I consider when deploying Kong (or any alternative API gateway) in a multi-node Kubernetes production cluster? How would you approach planning for its scalability?
1 Answer
Based on my experience with Kong and Ingress, I'd recommend sticking with the kong/kong setup since it offers more flexibility, even though it involves more components. It's best to keep the data plane, control plane, and controller installations separate and customized for each one. This separation helps in managing them better in a production environment.
Is a separate controller really necessary? I thought having just the control and data planes would suffice if I'm not using the gateway API features. I was planning to use Kong Manager for setup.