[Call for feedback] Templates for deployments

Users
1

Hello MAAStronauts,

We’d love your input on some ideas we’re exploring to improve how MAAS configures and deploys machines, especially at scale.

Right now, deploying a machine in MAAS involves:

  • Configuring network interfaces

  • Setting up the storage layout

  • Then initiating the deployment

While this works well in small environments, it becomes repetitive and error-prone when dealing with large datacenters. Manually configuring each machine doesn’t scale well.

To address this, we’re considering adding support for some sort of “machine classes”, a way to group and manage heterogeneous hardware consistently. We’re considering also “deployment templates”, which would define different network configurations and storage configurations for each class. (names are still subject to change.).

For example, if you have 100 servers with the same hardware (i.e. also the same amount of disks with the same capacity), you could create multiple storage templates with different layouts or RAID configurations. When deploying a machine from that class, you would simply choose the template you want to apply. There will be no need to configure each server manually.

We’re also exploring support for three primary deployment workflows:

  • Deploy a specific machine: choose a known machine and apply a template.

  • Deploy from a machine class: request a server from a specific class (e.g., any high-memory node), and MAAS will automatically pick and configure one using the selected template.

  • Deploy a machine with some minimum requirements that can accomodate the selected templates: request a server that has some minimum caracteristics (for example, the amount of memory) and use the selected templates to deploy it.

We’re reaching out to gather your use cases, feedback, and ideas. Some basic questions to initiate the conversation are:

  • What challenges do you face when deploying machines at scale?

  • Would machine classes and templates help streamline your workflow?

Feel free to reply here, send me a PM, or let’s set up a quick call if you’d prefer. Your insights will help shape this feature!