Productive K3S Core Modes

productive-k3s-profiles relies on the explicit installation modes exposed by productive-k3s-core.

The modes

• single-node: one-node bootstrap on a single machine
• server: initialize or join a K3S server node
• agent: join an existing K3S server
• stack: install cluster-level components after the cluster exists

Why they matter here

The infrastructure repository needs to assemble clusters predictably after machine provisioning or host discovery.

That split makes the orchestration model explicit:

1. create or target machines first
2. establish the cluster shape second
3. install the shared stack last

How scenarios consume them

• multipass: explicitly uses server, agent, and stack
• onprem-basic: can exercise single-node for one host, or server, agent, and stack for multi-node layouts
• aws-single-node: currently packages a single-node public flow around the shared remote bootstrap layer

Notes

Note

The more explicit the bootstrap modes are in productive-k3s-core, the easier it is for infrastructure automation to stay understandable and testable.

Note

The optional experimental environment variable PRODUCTIVE_K3S_ENGINE can switch the base K3S installation backend between native and k3sup without changing the mode contract. The shared stack path remains unchanged.

Note

k3sup support is complementary and experimental. The product scope, supported matrix, and repository guarantees are still defined by Productive K3S itself, not by every workflow that k3sup could theoretically enable.