Documentationbreadcrumb arrow Grafana Cloudbreadcrumb arrow Knowledge graphbreadcrumb arrow Use the knowledge graphbreadcrumb arrow Grafana Assistant
Grafana Cloud

Use Grafana Assistant in RCA workbench to solve performance issues

Grafana Assistant in RCA workbench can help you accelerate root cause analysis by forming hypotheses quickly. It helps you diagnose and resolve performance issues faster by analyzing entities, timelines, and dependencies across your observability stack.

Grafana Assistant provides the following capabilities:

  • Intelligent root cause analysis: Analyzes entities, timelines, and dependencies to form hypotheses and identify root causes
  • Natural language troubleshooting: Query metrics, logs, traces, and profiles using plain English without writing complex queries
  • Contextual analysis: Uses the knowledge graph to provide AI models with context about your infrastructure and services
  • Automated workflows: Streamlines investigations with built-in command-based tools
  • Multi-signal correlation: Links related data across metrics, logs, traces, and profiles
  • Guided remediation: Suggests next steps and recommended actions based on its analysis

Grafana Assistant is also available throughout Grafana in other contexts, providing consistent AI-powered help across dashboards, drilldown, and other features. Refer to the Grafana Assistant documentation to learn more.

In RCA workbench, it has specialized capabilities for root cause analysis.

Before you begin

Before using Grafana Assistant in RCA workbench:

  1. Add entities: Add relevant entities (for example, services and pods) to RCA workbench for the assistant to analyze
  2. Narrow the time range: Focus on the specific incident window for more accurate results
  3. Verify data sources: Ensure that metrics, logs, and traces are connected and accessible

Open Grafana Assistant

After you’ve added entities to RCA workbench:

  1. Click Analyze RCA Workbench in the bottom-right corner or open the chat interface.
  2. Allow the assistant to complete its initial analysis before asking questions.

Work with Grafana Assistant

Learn to use Grafana Assistant features to enhance root cause analysis.

Reference entities with @ mentions

When asking questions about specific entities, use the @ command to explicitly reference them. This ensures the assistant focuses its analysis on the exact entities you want to investigate. For example: @frontend-service what is causing the latency spike?

Ask natural language questions

You can ask questions in plain English. Grafana Assistant automatically queries metrics, logs, and traces as needed.

You can ask complex, multi-step questions in plain English, such as:

  • Show the request rate of all services that connect to my service
  • What entities are experiencing errors in the production namespace?
  • Compare CPU usage between @api-service and @payment-service

The assistant automatically executes multiple tool calls as needed to answer your questions.

Grafana Assistant capabilities in RCA workbench

When used in RCA workbench, Grafana Assistant gains context-aware capabilities:

  • Entity health analysis: Examines entity properties, relationships, and health anomalies
  • Timeline explanation: Interprets and explains the events shown in your timeline view
  • Graph exploration: Helps navigate entity relationships and dependencies
  • Pattern correlation: Identifies common issues across multiple entities
  • Infrastructure memory: Uses knowledge about your system topology and service groups
  • Workbench management: Can help add or remove entities, adjust filters, and refine your investigation scope

Grafana Assistant understands your current workbench context, including:

  • Selected entities and their relationships
  • Active time range
  • Applied filters (environment, region, namespace)

Best practices to use Grafana Assistant

  • Be specific about time ranges: Narrow your investigation to the specific incident window. This improves analysis speed and relevance.
  • Start with key entities: Add the most critical or problematic entities first. You can always add connected entities later based on assistant recommendations.
  • Use @ mentions for precision: When investigating specific services or components, explicitly reference them with @ to get focused analysis.
  • Follow suggested next steps: The assistant highlights recommended actions based on its analysis. These are tailored to your specific investigation context.
  • Iterate and refine: Use the assistant conversationally and ask follow-up questions, request clarification, or pivot your investigation based on findings.

Troubleshoot Grafana Assistant

  • Assistant isn’t responding or seems stuck:
    • Verify your Grafana Assistant settings and confirm that you’ve accepted the terms and conditions
    • Check that you have entities added to the workbench
    • Ensure your time range isn’t too broad. Narrow to specific incident window.
    • You have exceeded the token limit in the context window for a single session
  • Analysis is taking too long:
    • Reduce the number of entities in your workbench
    • Use a more specific time range
    • Be more specific in your questions using @ entity mentions
  • Recommendations seem irrelevant:
    • Verify the correct entities are selected
    • Check your filters (environment, region, namespace)
    • Provide more context in your questions about what you’re investigating