Grafana Launches AI-Powered Assistant to Diagnose Database Slowdowns in Real-Time

Grafana Labs today unveiled a new AI-driven assistant integrated directly into its Database Observability tool, promising to slash the time it takes to pinpoint and resolve performance bottlenecks. The feature, announced Tuesday, allows engineers to instantly analyze slow queries, wait events, and schema issues without manually assembling context from separate systems.

“We’re not just giving you more dashboards—we’re giving you an AI teammate that works with your actual data,” said Tom Wilkie, CTO of Grafana Labs. “It queries your Prometheus and Loki sources live, understands your schema, and then tells you exactly what to fix.”

The assistant targets the most frustrating part of database troubleshooting: translating raw metrics like P99 latency spikes or wait events (e.g., wait/synch/mutex/innodb) into actionable diagnoses. Instead of copying SQL into a separate chatbot, the tool runs analyses against the same time window and table metadata already loaded in the Grafana Cloud interface.

“Visibility alone isn’t enough—you need context-aware guidance,” added Sarah Novotny, a database reliability engineer at Grafana. “The assistant doesn’t guess; it uses live execution plans and index definitions to explain why a query suddenly slowed.”

How It Works

Each query tab now includes purpose-built analysis buttons designed by database engineers, not generic AI prompts. Clicking “Why is this query slow?” triggers an automated investigation that pulls real-time RED metrics (Rate, Errors, Duration), examines wait events, and compares rows examined versus rows returned.

In a demonstration, the assistant discovered that a query’s duration spiked because it examined 50 times more rows than it returned—a classic table scan waste. It also flagged that CPU time remained healthy while wait events consumed 40% of execution time, pointing directly to I/O contention.

“The names are intimidating—like wait/synch/mutex/innodb—but the assistant translates them into plain English and says, ‘This is a locking problem, here’s how to reduce lock time,’” Novotny said.

Privacy and Security

Grafana emphasized that query text and schema metadata are used only for the current analysis and are never stored or used for model training. “We built this to respect data security,” Wilkie noted. “Your sensitive SQL stays in your environment.”

Background

Database performance monitoring has long relied on dashboards and manual correlation of metrics, logs, and traces. Engineers often juggle multiple tools—Prometheus for metrics, Loki for logs, and separate SQL explain tools—to form a complete picture. The Grafana Cloud Database Observability product already aggregated these signals, but left the diagnosis step to the user.

The new assistant fills this gap by automating the correlation step. Pre-defined prompts for common issues—such as slow queries, degraded throughput, or bad indexes—guide users along best-practice paths. AI models are fine-tuned on database knowledge, not generic internet data, ensuring advice is specific to MySQL, PostgreSQL, and other supported databases.

Early beta testers reported a 50–70% reduction in time spent troubleshooting routine slowdowns, according to Grafana. The tool integrates with existing alerting workflows, allowing SREs to trigger an assistant analysis directly from a PagerDuty notification.

What This Means

For site reliability engineers and database administrators, the assistant shifts the focus from data gathering to decision-making. Instead of spending 15 minutes figuring out that a mutex wait is causing a bottleneck, they can act immediately on recommended fixes—such as adding an index, rewriting a join, or adjusting a configuration parameter.

“This brings AI-assisted troubleshooting from a general-purpose chat into a purpose-built engineering tool,” said industry analyst Mike Gualtieri of Forrester Research. “Context is everything in observability, and Grafana is betting that embedding AI directly into the context will outperform any standalone chatbot.”

The integration is available today for Grafana Cloud customers with Database Observability enabled. Pricing includes the standard Cloud usage costs with no additional charge for the assistant feature during its initial rollout.

For more details on how to get started, visit the official documentation.

Tags:

Grafana Launches AI-Powered Assistant to Diagnose Database Slowdowns in Real-Time