Should I Host Everything in My Cluster or Use Managed Services?

For databases like PostgreSQL and RabbitMQ, we've opted to run those on virtual machines instead of clusters. Our database administrators aren't very Kubernetes savvy, and getting them to debug pods was just too challenging. Also, just a heads up, B1S nodes on Azure are burstable and may not be suitable for production use.

In my experiences, we use managed services whenever possible. If the cloud provider offers something beneficial, like Kafka or Flink, we take advantage of it. But for databases, relying on the cloud provider’s services is usually more effective, especially for complex needs.

If you're considering running a database on Kubernetes, I recommend something like CNPG, but be cautious because Kubernetes isn't really designed for databases. Managed database services take a load off your shoulders, and if cost isn’t a huge concern, they are definitely the way to go.

It really varies based on your traffic and whether your workload is constant or fluctuating. If you have stable traffic, self-hosting could work well.

Cost is a big factor here. Running everything in your cluster is usually cheaper when scaling up. Managed services are convenient but can get pricey. Plus, using something like Crunchy for PostgreSQL is solid. Just keep in mind that self-hosting does come with some management responsibilities. If something goes wrong, like data loss or downtime, that could be a big issue. In my project, we run ephemeral caches on k8s, but for critical data, we prefer managed services from cloud providers.