Observing Agentic/MCP Traffic & Keeping Costs Low
summary
Installs and configures agent gateway on Kubernetes, creates three gateways and backend CRDs that route to an Anthropic/Claude model by rewriting paths, then curls each ALB IP to generate traffic and metrics. Installs Cube Prometheus, port-forwards Prometheus and Grafana, adds pod monitors scraping /metrics every 30s, imports a custom Grafana dashboard showing cost per gateway and token usage and demonstrates the cost breakdown while staying within budget.
AI-written summary · about
chapters (9)
- 0:00Introduction to Conf 42 DevOps 2026
- 0:43Setting Up Agent Gateway
- 0:53Choosing and Configuring LLM Provider
- 1:23Creating and Configuring Gateways
- 3:34Testing Gateway Configuration
- 4:34Installing and Configuring Q Prometheus
- 5:32Setting Up Grafana Dashboards
- 7:01Monitoring and Analyzing Costs
- 8:18Conclusion and Final Thoughts
related talks
28mFrom Manual to Managed: Prometheus Agent Deployment at Scale
24mCost-Efficient Multi-Cluster Monitoring with Prometheus, Grafana & Linkerd
24mPractical Kubernetes Monitoring with Prometheus
41mFrom Chaos to Order: How to Konnect, secure and observe [...]
35mUnlocking Kubernetes Observability: Secure, Tenant-Cen...
25mKubernetes Ate the Budget: How We Made Cost a Platform Responsibi...
24mBanking Observability at Scale
37mMonitoring Kubernetes with Prometheus