Add node autoscaling

autoscaling variable can be set when using module specifying both min and max node counts:

module "cluster" {
  source = "git::ssh://git@gitlab.developers.cam.ac.uk/uis/devops/infra/terraform/gke-cluster.git"

  project  = local.project
  location = "europe-west2"

  machine_type = "n1-standard-1"

  autoscaling = {
    min_node_count = 1
    max_node_count = 3
  }
}

Behaves as normal without having autoscaling variable set.

Also added functionality to be able to specify basic node_count (default 1) if not using autoscaling.

Closes #10 (closed)

changed milestone to %Sprint 95

how does it autoscales?

If the requested resource for a pod doesn't match the available resource for the cluster, a new node is added to schedule the pod. This only really works well if we make sure to annotate the deployments with sensible CPU and memory requirements which we found was somewhat "empirical" with GitLab.

This only really works well if we make sure to annotate the deployments with sensible CPU and memory requirements

Which I believe is this one: https://gitlab.developers.cam.ac.uk/uis/moodle/moodle-cloud/deployment/-/issues/32

Must admit that I've now got node autoscaling, cluster auto-provisioning and horizontal pod autoscaling all spinning wildly in my head and am not sure which is the way forward.

This MR is still virtuous and worth merging IMO.

You've now graduated from terraform wizard to sorcerer :).

Feel more on a par with Rincewind

Oook.

approved this merge request

unapproved this merge request

added 1 commit

• 118ff3ff - Apply suggestion to variables.tf

Compare with previous version

Doing a quick test with my favourite whipping boy: Moodle.

You'll have go fight for lock on terraform state...

:) - I'm just doing a plan to make sure the right values get set.

changed the description

approved this merge request

This correctly sets the appropriate values on the cluster. Making use of them is an exercise for the deployment :).

mentioned in commit f8c32143

merged