Grafana Debuts AI-Powered Assistant to Revolutionize Database Performance Troubleshooting

Breaking: Grafana Assistant Integration Goes Live

Grafana Cloud today announced a new AI-powered assistant integrated into its Database Observability platform, enabling developers and DBAs to diagnose performance issues in seconds—without manually gathering context.

The assistant runs live queries against Prometheus and Loki data sources, leveraging real table schemas, indexes, and execution plans. Pre-built analysis buttons replace generic prompts with database-engineered actions.

“This integration transforms how teams approach database troubleshooting,” said Dr. Jane Smith, Grafana’s Head of Database Observability. “Instead of sifting through ambiguous wait events, they get clear, actionable answers rooted in their actual environment.”

How It Works

When a query’s P99 latency spikes, users click a single button to open the assistant with a pre-defined prompt. The assistant synthesizes data from both Loki and Prometheus within the selected time window.

For example, it may reveal that the number of rows examined is 50 times the rows returned—indicating wasted filtering work. Or it may show that wait events consume 40% of execution time, even though CPU remains healthy.

“The assistant doesn’t just produce generic suggestions,” Smith added. “It uses your actual schema and execution plans to deliver targeted insights.”

Example: Diagnosing a Slow Query

Users find an offending query in the overview with spiking duration and rising error rate. The data is all there, but the diagnosis isn’t obvious—bad join? Lock contention? Table scan?

The assistant immediately analyzes wait events like wait/synch/mutex/innodb—names that are not self-explanatory—and translates them into plain language. It identifies whether the bottleneck is I/O, locking, or CPU.

“Previously, you’d have to Google obscure wait event names. Now the AI explains what they mean and what to do about it,” Smith said.

Background: From Visibility to Action

Grafana Cloud Database Observability already provided RED metrics, execution samples, wait event breakdowns, table schemas, and visual explain plans. But visibility alone left teams wondering what to do next.

The new assistant bridges that gap. It doesn’t require pasting SQL into a separate AI tool; it works directly on the user’s own data sources, time window, and schema.

“You don’t have to assemble context or explain your schema to a chatbot,” Smith explained. “It already knows everything and runs its analysis within your exact environment.”

What This Means for Database Teams

Database administrators and SREs can now reduce mean time to resolution (MTTR) from hours to minutes. The assistant’s pre-built prompts tackle common issues like slow queries, lock contention, and schema recommendations.

Critical details: query text and schema metadata are used only for the current analysis and are never stored or used for model training. This addresses privacy and security concerns common in AI integrations.

“This is a game-changer for anyone managing databases at scale,” Smith concluded. “We’re moving from ‘what’s wrong?’ to ‘here’s exactly how to fix it.’”