AI-powered Monitoring & Agentic Operations for Druid 37.0.0

[Unknown]

Hi everyone,

I'm currently running an Apache Druid 37.0.0 cluster with 20+ nodes and am looking to improve our monitoring and operational workflows using AI.

I'm interested in learning how others in the community are approaching AI-assisted monitoring and operations for Druid. Specifically:

1. Monitoring

   • Which Druid and infrastructure metrics do you consider the most critical to monitor?

   • If you're using Prometheus, which metrics and alerting rules have proven the most valuable in production?

   • Are there any dashboards (Grafana or otherwise) that you would recommend as a starting point?

2. AI / LLM Integration

   • Has anyone integrated Druid monitoring with AI agents (OpenAI, Claude, Gemini, etc.) for anomaly detection, root-cause analysis, or operational assistance?

   • Have you found this useful in production, and if so, what architecture or tooling are you using?

3. MCP (Model Context Protocol)

   • Are there any mature MCP servers or integrations that work well for infrastructure operations?

   • I'm particularly interested in capabilities such as:

     • Executing controlled operational actions (for example, restarting Druid services after approval)

     • Reading Prometheus metrics

     • Querying Grafana dashboards

     • Inspecting logs

     • Diagnosing Druid issues

   • If you've built something similar, I'd appreciate hearing about your experience and any lessons learned.

I'm looking for practical production experiences, recommended tools, architectures, or open-source projects that have worked well.

Thanks in advance!