How Turtles Are Made

How many turtles does it take?

We have lots of jobs and services to run. I think we’re gonna need a lot of machines to run them on.

We have lots of workloads and machines. I think we’re gonna need a scheduler to run them.

We have lots of different types of workloads to run with this scheduler. I think we’re gonna need an orchestration API to manage it.

We have lots of scripts using the orchestration API. I think we’re gonna need to automate these controllers.

We have lots of controllers using the orchestrator. I think we’re gonna need to make it highly available.

We have lots of orchestrator components. I think we’re gonna need a better installer.

We have lots of changes to roll out to the orchestrator components. I think we’re gonna need an upgrader.

We have an installer and upgrader. I think we’re gonna need to run them as a service so they can manage the full orchestrator lifecycle over time.

We have an orchestration cluster manager service. I think we’re gonna need to run it on a separate bootstrap machine.

We have lots of changes to roll out to the cluster manager. I think we’re gonna need a CLI to upgrade the bootstrap machine.

We have lots of workloads that need to be isolated or spread across multiple regions. I think we’re gonna need multiple clusters.

We have lots of clusters managed by the cluster manager. I think we’re gonna need to make it highly available.

We have lots of cluster manager components. I think we’re gonna need a better installer.