You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We want to split provisioner daemon and provisioners so they can run on separate workloads. Right now provisioner daemon unpacks the template and sends the directory path to provisioner, but that can't work if they're separate.
We need to refactor so that the provisioner daemon sends the template archive to the provisioner, and the provisioner unpacks it.
The text was updated successfully, but these errors were encountered:
What is the difference between the Provisioner and the Provisioner Daemon?
EDIT: I saw the RFC. I believe in the current implementation the Provisioner lives inside the Provisioner Daemon. The Provisioner Daemon has an internal memory queue for dispatching jobs. However the future goal is to split these.
Then you could have the actual Provision step be a different Job system, such as a Kubernetes Job, or a Lamda action, etc. So the daemon dispatches the jobs to the external job handling system (some cloud job queue thing likely)
What is the difference between the Provisioner and the Provisioner Daemon?
The Provisioner is the thing that actually executes code to provision workspaces, and it comes in different types. Right now we have Echo and Terraform. Echo is just for testing, but Terraform Provisioner is the thing that actually shells out to terraform.
The Provisioner Daemon is responsible for communicating with Coderd: acquiring, updating, and completing jobs, and then calling into the appropriate Provisioner to do the work. Right now the Daemon also unpacks the template archive, but the goal of this issue is to refactor so the Provisioner unpacks the template archive.
I believe in the current implementation the Provisioner lives inside the Provisioner Daemon. The Provisioner Daemon has an internal memory queue for dispatching jobs. However the future goal is to split these.
Then you could have the actual Provision step be a different Job system, such as a Kubernetes Job, or a Lamda action, etc. So the daemon dispatches the jobs to the external job handling system (some cloud job queue thing likely)
We want to split provisioner daemon and provisioners so they can run on separate workloads. Right now provisioner daemon unpacks the template and sends the directory path to provisioner, but that can't work if they're separate.
We need to refactor so that the provisioner daemon sends the template archive to the provisioner, and the provisioner unpacks it.
The text was updated successfully, but these errors were encountered: