Skip to content

Add AWS EKS API autoscaling (Prometheus + adapter) - #94

Draft
Erik Weathers (erikdw) wants to merge 1 commit into
brianvans/support-autoscaling-api-on-gkefrom
erikdw/api-autoscaling-aws-prom
Draft

Add AWS EKS API autoscaling (Prometheus + adapter)#94
Erik Weathers (erikdw) wants to merge 1 commit into
brianvans/support-autoscaling-api-on-gkefrom
erikdw/api-autoscaling-aws-prom

Conversation

@erikdw

Copy link
Copy Markdown
Contributor

Summary

  • Stacks on the GKE autoscaling branch: keep AutoscalingMetric for cloud=google, add in-chart Prometheus + prometheus-adapter for cloud=aws.
  • HPA uses the same CPU / event-loop ratio (0.4) / delay mean seconds (0.05) targets; AWS metric names come from the adapter (event_loop_*).
  • Unit tests cover Google vs AWS template selection and unsupported-cloud failure.

Test plan

  • ./test.sh (unit + render + lint) locally
  • Helm upgrade on erikdw-sandbox2 with cloud: aws and api.autoscaling.enabled: true
  • Confirm Prom scrapes :8001/metrics, custom.metrics has ELU/delay, HPA currents non-zero
  • Rebase/retarget onto main after the GKE PR merges

Made with Cursor

Stack on the GKE AutoscalingMetric path so cloud=aws scrapes health
/metrics and feeds HPA through custom.metrics, reusing the same
event-loop ratio/seconds targets.

Co-authored-by: Cursor <cursoragent@cursor.com>
@erikdw

Copy link
Copy Markdown
Contributor Author

Sync / retest (2026-08-10)

  • Already contains main (no rebase needed); still correctly stacked on GKE #95.
  • Re-validated on erikdw-sandbox2: Prom targets up on :8001/metrics, custom.metrics ELU/delay present (1m / 20m), HPA cpu: 0%/50%, 1m/400m + 1 more….
  • Helm-upgraded with current chart + nested api.autoscaling.* values schema.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant