Grafana Launches AI-Powered Assistant for Rapid Database Performance Troubleshooting

Grafana Labs today announced a new AI-driven assistant integrated into its Database Observability platform, enabling engineers to diagnose and fix slow SQL queries in minutes instead of hours.

“This integration transforms how teams hunt down database performance issues,” said Sarah Chen, VP of Product at Grafana Labs. “Instead of manually piecing together context across multiple tools, the assistant directly queries your actual data sources and provides actionable insights.”

Background

Database performance problems have long plagued developers and site reliability engineers. Traditional monitoring tools show metrics like P99 latency spikes or cryptic wait events such as wait/synch/mutex/innodb, but translating those into root causes requires deep expertise and manual analysis.

Grafana Cloud Database Observability already provided visibility into SQL queries with RED metrics, execution samples, wait event breakdowns, table schemas, and visual explain plans. However, visibility alone wasn’t enough.

New Assistant: How It Works

The new Grafana Assistant integration for Database Observability combines AI with the full depth of Grafana Cloud’s observability capabilities. When investigating a query, engineers open the assistant with a single click. It automatically leverages the same Prometheus and Loki data sources already in use, within the exact time window under analysis.

“The assistant doesn’t rely on a copy-pasted SQL snippet into a separate AI tool,” Chen explained. “It runs live queries against your actual Prometheus and Loki data, with real table schemas, indexes, and execution plans already loaded. That context is critical for accurate diagnosis.”

Each tab includes purpose-built analysis actions designed by database engineers, not generic prompts. Analyses are based on real database data and provide specific, actionable advice. Query text and schema metadata are used only for the current analysis and are never stored or used for model training.

Built-in Prompts for Common Issues

While users can still freely prompt the assistant chat, Grafana has added out-of-the-box AI buttons for tackling slow or degraded queries and getting change recommendations.

Example: Why Is This Query Slow?

An engineer spots a query whose duration is spiking and error rate climbing. Clicking into the query shows time-series performance data, but the diagnosis isn’t obvious. Is it a bad join, lock contention, or a table scan that was fine until data grew?

With one click, the assistant uses both Loki and Prometheus to query the selected time window and synthesizes a single health assessment. For instance, it might reveal that duration is spiking because the number of rows examined is 50 times the rows returned, meaning most work is wasted on filtering. P99 is 12 times the median, indicating an intermittent problem. CPU time is healthy, but wait events consume 40% of execution time.

Wait event names like wait/synch/mutex/innodb or io/table/sql/handler are not self-explanatory. The assistant understands them and explains: “During this wait, the database is physically….”

What This Means

For teams managing large-scale databases, the Assistant reduces mean time to resolution (MTTR) for performance incidents. It removes the need to manually correlate metrics, logs, and schema information, and instead provides a unified, data-driven analysis in seconds.

“Engineers spend far too much time context-switching between tools and trying to interpret cryptic database internals,” Chen said. “This integration lets them focus on fixing the issue, not hunting for clues.”

The Assistant is available now in Grafana Cloud Database Observability for all customers.