[spike] Support standing-up clusters in Terraform (for AWS) - #8642
[spike] Support standing-up clusters in Terraform (for AWS)#8642agoose77 wants to merge 9 commits into
Conversation
|
Merging this PR will trigger the following deployment actions. Support deploymentsNo support upgrades will be triggered Staging deployments
Production deployments
|
|
This was one of the reasons i moved us to using managed nodegroups! If this works, great :) But do timebox yourself, and make sure upgrades are possible |
|
Also big +1 for just using the terraform resources directly rather than using a module like https://registry.terraform.io/modules/terraform-aws-modules/eks/aws/latest. Our needs are reasonably simple so this is ok. |
525365c to
ab6687e
Compare
| role = aws_iam_role.node[0].name | ||
| } | ||
|
|
||
| resource "aws_launch_template" "core" { |
There was a problem hiding this comment.
This is setting up an unmanaged nodegroup, right? We should use https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/eks_node_group instead which should drastically simplify this
There was a problem hiding this comment.
This should be setting up managed node groups. We need to specify launch templates so that we can patch the kubelet config.
That said, I haven't tested this yet — I mainly tested that the plan looked reasonable — this was a short spike to build out the infra.
This PR recreates the neurohackademy cluster.
I generally dislike the need to run
eksctl— it's slow, and a separate workflow from our other clusters. I noticed #1924 today, and I think it still holds true.Therefore, I'm taking this opportunity to rebuild neurohackademy using Terraform.
To Do