feat: Discover Instance Type Memory Capacity

[jukie]

Fixes #5161

Description
This adds a new controller for managing a cache on the instancetype provider that stores the memory capacity overhead for each instance type by comparing the actual value after a Kubernetes Node gets registered with the cluster.
The cache is then implemented by NewInstanceType() when calculating memory capacity. If a cached value doesn't exist it will fall back to the existing logic of vmMemoryOverheadPercent.

The cache is keyed by a combination of instance type name and hash of nodeClass.Status.AMIs so this should always be accurate and when an AMI update is triggered will use the existing logic of calculating against vmMemoryOverheadPercent to ensure safe instance creation every time.

How was this change tested?
Suite tests and live environment

Does this change impact docs?

• Yes, PR includes docs updates
• Yes, issue opened:
• No

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[netlify]

✅ Deploy Preview for karpenter-docs-prod ready!

Name	Link
🔨 Latest commit	90d9c55
🔍 Latest deploy log	https://app.netlify.com/sites/karpenter-docs-prod/deploys/67107692a7a1e80008f75dd1
😎 Deploy Preview	https://deploy-preview-7004--karpenter-docs-prod.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[njtran]

Thanks for the PR! Looks really good!

[jukie]

CC @jmdeal (I think you were the one on the call today?) There's a negligible amount of variance when measuring overhead in Ki but I was unable to find any variance between the same instance type when measuring in Mi which my PR is doing.

Here's a simple script that can be used to verify: https://gist.github.com/jukie/df045af8fec68941f5d119044bf04aee

Instance Type: m6i.24xlarge
Variance detected in memory capacities:
 - 389990024
 - 389990016
 - 389990004
 - 389989996
 - 389990000
 - 389990004
 - 389989988
 - 389990000
 - 389990012
 - 389990012
 - 389989988
 - 389989980
 - 389990020

[jmdeal]

/karpenter snapshot

[github-actions]

Snapshot successfully published to oci://021119463062.dkr.ecr.us-east-1.amazonaws.com/karpenter/snapshot/karpenter:0-7194ce607171bb52280313c434a574b616877c85.
To install you must login to the ECR repo with an AWS account:

aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 021119463062.dkr.ecr.us-east-1.amazonaws.com

helm upgrade --install karpenter oci://021119463062.dkr.ecr.us-east-1.amazonaws.com/karpenter/snapshot/karpenter --version "0-7194ce607171bb52280313c434a574b616877c85" --namespace "kube-system" --create-namespace \
  --set "settings.clusterName=${CLUSTER_NAME}" \
  --set "settings.interruptionQueue=${CLUSTER_NAME}" \
  --set controller.resources.requests.cpu=1 \
  --set controller.resources.requests.memory=1Gi \
  --set controller.resources.limits.cpu=1 \
  --set controller.resources.limits.memory=1Gi \
  --wait

[jukie]

@jmdeal could you take another look please

[coveralls]

Pull Request Test Coverage Report for Build 11377237253

Details

• 58 of 83 (69.88%) changed or added relevant lines in 6 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.2%) to 82.883%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
pkg/controllers/controllers.go	0	1	0.0%
pkg/operator/operator.go	0	1	0.0%
pkg/providers/instancetype/instancetype.go	43	49	87.76%
pkg/controllers/providers/instancetype/capacity/controller.go	10	27	37.04%

Totals
Change from base Build 11372026312:	-0.2%
Covered Lines:	5641
Relevant Lines:	6806

---

💛 - Coveralls

[jukie]

Fixed the conflicts

[jukie]

Any other changes needed here @jmdeal?

[jmdeal]

One nit, and I'm going to let E2Es run. Otherwise LGTM.

/karpenter snapshot

[jmdeal]

/karpenter snapshot

[github-actions]

Snapshot successfully published to oci://021119463062.dkr.ecr.us-east-1.amazonaws.com/karpenter/snapshot/karpenter:0-7548492412f9ec41de90fd048119ccc0ba7d0141.
To install you must login to the ECR repo with an AWS account:

aws ecr get-login-password --region us-east-1 | docker login --username AWS --password-stdin 021119463062.dkr.ecr.us-east-1.amazonaws.com

helm upgrade --install karpenter oci://021119463062.dkr.ecr.us-east-1.amazonaws.com/karpenter/snapshot/karpenter --version "0-7548492412f9ec41de90fd048119ccc0ba7d0141" --namespace "kube-system" --create-namespace \
  --set "settings.clusterName=${CLUSTER_NAME}" \
  --set "settings.interruptionQueue=${CLUSTER_NAME}" \
  --set controller.resources.requests.cpu=1 \
  --set controller.resources.requests.memory=1Gi \
  --set controller.resources.limits.cpu=1 \
  --set controller.resources.limits.memory=1Gi \
  --wait

[jukie]

I think the e2e's need a retry after committing your suggestion but the prior run appeared to pass.

[jmdeal]

No need to rerun, the only changes since they last ran was the comment nit and some docs changes you merged in. This LGTM 🚀.

[jukie]

Hi @jmdeal I saw 1.0.7 was released but this wasn't included. Will this make it into the next one?

[jmdeal]

We weren't planning on including this in a patch release, that's typically reserved for critical bug and security fixes. The plan would be to include this in the next minor version release, v1.1.0.

[jukie]

Thanks!