Reference · built on requestOperator by FibricIT & reliability

Cost Anomaly

Matches a jump in cloud spend to the deploy that preceded it and proposes a rollback or a rightsizing issue for approval.

About

Cloud bills arrive after the fact. A deploy that doubles a queue's consumers or turns on verbose logging shows up as a line item a day later, and nobody connects the two. Cost Anomaly reads the AWS Cost and Usage Report from the S3 bucket it is delivered to, the Cloud Billing export tables in BigQuery, and the EstimatedCharges metric in Amazon CloudWatch. It keeps a baseline per service, account, and usage type, and lines each jump up against the deployments recorded in GitHub for the same window.

When spend leaves its baseline, it names the service and the deploy that landed just before, and proposes what to do: request a deployment of the previous ref, or open an issue that lists the resources to rightsize. You approve each proposal. Nothing is redeployed or resized on its own.

This is a reference listing. It documents what Fibric would read from Cost Anomaly and what it could propose, based on the vendor's published interfaces. Fibric builds it under a managed deployment when you request it; selecting it here installs nothing.

Inputs

  • AWS Cost and Usage Report files as they land in your S3 bucket, with lineItem columns by product, usage type, and operation
  • Cloud Billing standard and detailed usage cost exports in BigQuery, with project, service, SKU, labels, cost, and credits
  • The EstimatedCharges metric in the AWS/Billing namespace of Amazon CloudWatch, by ServiceName, stored in US East (N. Virginia)
  • Deployments in GitHub with ref, sha, environment, and created_at, and their outcomes from deployment_status events
  • Refund and credit lines that AWS adds after a month closes, so a corrected bill does not read as a drop
  • Approve and decline responses on the proposals it posts to Slack

Proposed actions

  • Target capability: propose a deployment of the previous ref through the GitHub Deployments API when a spend jump follows a deploy
  • Target capability: propose a GitHub issue in the owning repository that lists the resources, usage types, and daily delta to rightsize
  • Target capability: propose a comment on the pull request or issue that shipped the change, showing the line items that moved
  • Target capability: propose the anomaly as a Slack message with the baseline, the delta, and approve and decline buttons

Proposed actions are target capabilities. Every action runs propose-first and needs a validated deployment and the appropriate permissions.

What you can build

  • Roll back the deploy that doubled the bill

    A release turned on per-request logging. The next report shows the logging service's usage type well above its baseline, shortly after a production deployment. Cost Anomaly proposes a deployment of the previous ref and posts it to Slack for the release owner.

    With Amazon S3, GitHub, Slack

  • Rightsize a project that grew quietly

    The BigQuery export shows one project's compute SKU rising week over week with no matching deploy. The operator proposes an issue listing the SKUs and the daily delta so the owning team can resize.

    With BigQuery, GitHub

  • Catch a jump before the report lands

    EstimatedCharges for one service climbs above its usual slope mid-month. Cost Anomaly flags it in Slack with the CloudWatch series and holds the rollback proposal until the next report confirms the line items.

    With Amazon CloudWatch, Slack

Requirements

  • A Cost and Usage Report delivered to an S3 bucket the S3 connector can list and read
  • Cloud Billing export to BigQuery enabled on the billing account, with a dataset the BigQuery connector can query
  • Receive Billing Alerts enabled in the AWS payer account, so EstimatedCharges reaches CloudWatch
  • A GitHub App installation with read access to deployments and write access to issues on the repositories that deploy
Authentication
Cost Anomaly holds no keys of its own. It reads billing exports and deploy records through the S3, BigQuery, CloudWatch, and GitHub connectors you connect, and posts proposals through Slack.

Limits

  • Billing lags. The report updates at least once a day and Cost Explorer trails usage by up to 24 hours; jumps surface a day late.
  • Report updates within a month are estimates. Charges change until AWS finalizes the bill, and refunds arrive in separate files after the month closes.
  • Cloud Billing exports carry resource-level detail only in the detailed usage cost export; the standard export attributes cost to project and SKU.
  • It proposes a deployment request and an issue. It does not resize instances, change autoscaling, or delete resources.

Access and pricing

Reference listing. Fibric builds the operator under a managed deployment when you request it. Your quote covers the build, capabilities, usage, and support.

Request Cost Anomaly ↗

Questions and answers

What does a person approve?
Each proposal is one action: a deployment request for the previous ref, an issue for rightsizing, or a comment on the change. It shows the baseline, the delta by service and usage type, and the deploy it is matched to. You approve, edit, or dismiss it.
What record does it leave?
A receipt per anomaly: the report rows or export rows behind it, the baseline it was measured against, the deploy it was matched to, who approved which action, and how to undo it. Dismissed proposals are kept with the reason you give.
Does it change anything on its own?
No. It reads exports and deploy history and writes only after approval, and then only through the GitHub and Slack connectors. It never calls the cloud provider to stop, resize, or delete a resource.
Ask about Cost Anomaly

Ask about the capabilities and requirements in this listing.

For project-specific requirements, contact Fibric.