potiuk opened a new pull request, #74: URL: https://github.com/apache/airflow-ci-infra/pull/74
On 2026-10-26 the RunsOn stack ran 920 on-demand vCPUs against 48 on Spot, because its Spot circuit breaker stayed tripped (fixed in #73). Nothing reported it until jobs started queuing on the full on-demand quota. `runs-on-usage-report.yml` runs every Monday. It compares the account's on-demand and Spot EC2 vCPU-hours in us-east-2 over the last 7 days (`AWS/Usage` `ResourceCount`, classes `Standard/OnDemand` and `Standard/Spot`). Above 30% on-demand, it keeps one issue open in this repository and comments each week's numbers on it. It closes the issue once the share is back under 30%. It fails, rather than reporting, when CloudWatch returns incomplete results or no usage. The numbers are account-wide, so other EC2 instances in the region count too. It authenticates through a new role, `runs-on-usage-report`, in `terraform/runs-on-access`. Only workflows on `main` can assume it, and it can only call `cloudwatch:GetMetricData`. That module is applied by hand by an account admin, so the workflow fails until that apply is done. Checks run locally: prek hooks (including zizmor) and shellcheck of the workflow script. I checked the metric dimensions against the account's CloudWatch metrics. Terraform was not available locally, so `terraform validate` of `runs-on-access` was not run. --- ##### Was generative AI tooling used to co-author this PR? - [X] Yes — Claude Opus 5 Generated-by: Claude Opus 5 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
