Zum Inhalt springen
NEUMarktplatz und dein Team: Verkaufe, was du baust, und bekomm Pläne und Änderungen, die getestet sind, bevor du sie siehst. Neuigkeiten ansehen

Bibliothek / Monitoring, Betrieb & Störungen

Monitoring, Betrieb & Störungen

AWS- und GCP-Kostenanomalien mit GPT-4o-mini erkennen und Slack benachrichtigen

Läuft jeden Morgen, vergleicht die AWS- und GCP-Ausgaben von gestern mit der jüngeren Historie, markiert Spitzen über Dollar-, Prozent- und z-Score-Schwellenwerte und sendet KI-zusammengefasste Warnungen an Slack.

Was sie bewirkt

## 💸 FinOps – Cloud Cost Anomaly Agent (AWS + GCP) For engineering and finance teams who find out about a runaway NAT gateway, a forgotten GPU instance or a BigQuery job gone wild when the monthly bill lands, weeks after the money is spent. ### How it works Every morning the workflow pulls yesterday's spend by service and region from AWS Cost Explorer and your GCP billing export. It compares each line with the previous day and the past week, and flags only real spikes that pass a dollar floor, a percentage jump and a z-score. For each AWS spike it pulls recent CloudTrail changes in that region. GPT-4o-mini combines them with the usage-type breakdown to suggest the most likely cause. Slack gets an alert with the cost, monthly impact, probable cause and two buttons: mark it expected, or ignore it for seven days. Every anomaly and decision is logged in Google Sheets. ### Setup steps 1. Create a Google Sheet with an Anomaly_Log tab. 2. Add AWS credentials with ce:GetCostAndUsage and cloudtrail:LookupEvents, and Google OAuth2 with BigQuery read for GCP. 3. Connect Slack and OpenAI. 4. In Set Cost Agent Config, turn each cloud on or off and set the GCP project, billing table, thresholds and your n8n base URL (used for the buttons). 5. Run once manually, then activate.

Die Schritte

  1. Every Day at 9 AM IST
  2. Set Cost Agent Config
  3. Read Anomaly Log Sheet
  4. AWS Enabled?
  5. Get AWS Cost by Service & Region
  6. Get AWS Cost by Usage Type
  7. Normalize AWS Costs
  8. GCP Enabled?
  9. Query GCP Billing Export
  10. Normalize GCP Costs
  11. Detect Day-over-Day Anomalies
  12. Loop Over Anomalies
  13. Build Run Summary
  14. Post Run Summary to Slack

Insgesamt 29 Schritte.

Ähnliche Automatisierungen