Documentationbreadcrumb arrow Grafana Cloudbreadcrumb arrow Platformbreadcrumb arrow Knowledge graphbreadcrumb arrow Introduction
Grafana Cloud

Introduction to the knowledge graph

Observability solutions add the parts of your system they monitor to the knowledge graph, and insights about the health of those parts. Open an entity to see its insights and its metrics, logs, traces, and profiles together.

The knowledge graph is included in every stack. Navigate to Observability > Entity catalog. If the catalog shows Get started, refer to Get started with the knowledge graph.

Entities and insights typically add 5% to 10% more active series. There is no separate charge. Refer to Knowledge graph pricing.

What problems does it solve?

Modern cloud environments are complex. A single user-facing issue might involve dozens of services, databases, and infrastructure components. Traditional monitoring approaches require you to:

  • Write queries to find problems hidden in your data
  • Manually correlate metrics, logs, and traces across multiple tools
  • Understand every service and dependency in your environment
  • Create and maintain hundreds of alert rules

The knowledge graph brings those services, databases, and infrastructure components together with their health, so you can start from the entity instead of a query.

Grafana Assistant uses those entities and relationships when you ask about your environment. Agents connected with the gcx CLI can query the same relationships.

Core concepts

The knowledge graph is built on entities, context-sensitive dashboards, and insights that work together to surface issues and guide troubleshooting.

Entities

Entities represent the components in your environment. Observability solutions add these entities to the knowledge graph:

  • Services - Microservices instrumented with OpenTelemetry or other APM tools
  • Pods - Kubernetes workload units
  • Nodes - Kubernetes cluster nodes or virtual machines
  • Databases - MySQL, PostgreSQL, Redis, and other data stores
  • Infrastructure - Load balancers, clusters, namespaces
  • Frontend applications - Browser apps monitored with Frontend Observability

Each entity has properties (environment, version, cluster) and relationships to other entities (a service depends on a database).

These solutions add entities from the telemetry they send:

Custom entities

Refer to Manage entities and relations to group discovered components, such as the services, Pods, and databases that belong to one product. You can attach insights to that group.

In Grafana Assistant, refer to Use Knowledge Graph mode to inspect existing connections and write the rules.

Coverage in preview

The following connections are in preview:

  • Synthetic checks (public preview). Refer to Connect Synthetic Monitoring checks to the knowledge graph to show a check next to the service it tests.
  • Microsoft SQL Server (private preview). Database Observability can add SQL Server databases to the knowledge graph.
  • AI agents (private preview). Agent Observability is generally available. Adding those agents to the knowledge graph is in private preview.
  • Service ownership (private preview). Service Center can show the owning team, the on-call schedule, and runbooks for a component.
  • Operational context (private preview). Grafana IRM can attach notes from an earlier investigation, including what the team tried and what happened next, to the related entities.

Context-sensitive dashboards

When you click on an entity, you see dashboards tailored to that entity type. Services show RED metrics (Request rate, Error rate, Duration). Infrastructure shows CPU, memory, and disk usage. Logs, traces, and profiles are pre-filtered to that entity.

Insights: The connective tissue

Insights are the core of how the knowledge graph helps you troubleshoot. They’re automatically generated alerts that track different types of issues:

  • Saturation - Resources approaching limits (CPU at 90%, disk filling up)
  • Amend - Configuration changes that might cause issues (deployments, scaling events)
  • Anomaly - Unusual patterns in traffic or behavior (request rate spike, latency increase)
  • Failure - Critical system failures (Pod crashes, controller errors)
  • Error - Request failures and latency threshold breaches (500 errors, slow responses)

When you see a service with a red indicator in the entity catalog, that’s an insight telling you something’s wrong. Click into it, and you’ll see exactly which type of issue is occurring, when it started, and what signals support it.

How insights and signals work together

Insights don’t replace your metrics, logs, and traces; they enhance them. An insight points you to a problem, then you drill into the underlying signals to understand why. For example:

  1. Insight: “Latency threshold breach” on checkout-service
  2. Click to view entity details
  3. See latency spike on dashboard
  4. Click the Traces tab to see which operations are slow
  5. Identify bottleneck in database query

This workflow takes seconds instead of minutes.

Key features

The knowledge graph provides several features to help you monitor and troubleshoot:

Entity catalog

Your primary entry point for monitoring. The entity catalog shows all discovered entities with their health status, insights, and key metrics. Filter by entity type, environment, or insight category to quickly find what needs attention.

Used in these workflows:

Entity graph

The entity graph provides a visual network view of relationships between your services and infrastructure.

RCA workbench

Add entities to RCA workbench to investigate incidents involving multiple components. See insights firing across different entities on a unified timeline, which allows you to spot which issues started together and identify root causes.

Used in these workflows:

Instrumentation quality

The instrumentation quality report identifies telemetry gaps that limit knowledge graph capabilities. It checks each service for missing metrics, incomplete labels, and configuration issues, and provides actionable recommendations to fix them. Use it when entities are missing or incomplete.

Grafana Assistant

Grafana Assistant uses entities and relationships from the knowledge graph when you ask about your environment. Query those relationships with Cypher in the knowledge graph data source. Agents connected with the gcx CLI can run the same queries.

Used in these workflows:

Infrastructure memory

Infrastructure memory is a capability of the knowledge graph. Memories summarize the services, infrastructure, and relationships in your stack. Grafana Assistant uses those summaries with the entities in the knowledge graph.

What makes it different?

Unlike traditional APM or monitoring tools:

  • Builds on telemetry you already send - Observability solutions add entities from telemetry already in your stack
  • Insights without writing queries - The entity catalog and insights surface issues without a query for each one
  • Unified workflow - Metrics, logs, and traces in one place, pre-filtered by entity
  • Included in Grafana Cloud - Available in every stack from Observability > Entity catalog. Refer to Knowledge graph pricing for the active series this adds
  • Used by Grafana Assistant - Assistant answers questions about your environment using entities and relationships from the knowledge graph

Next steps