Skip to content
Last9
Book demo

Discover Databases

Database-centric view of datastores discovered from application traces and supported database receivers

The Databases feature in Discover gives you a database-centric view of datastores Last9 detects from application database spans and supported database receivers. Instead of starting from a service and drilling into its database calls, start from the database itself to explore query activity and the services associated with it.

Discover Databases list view

Application database spans identify datastores through attributes such as db.system and net.peer.name. Supported database receivers can also discover databases and report database-side query activity. Receiver support and available views depend on the database type and the Last9 API and UI releases enabled in your workspace.

Prerequisites

For service-level database calls, your services need OpenTelemetry instrumentation that records database client spans. Most language auto-instrumentations do this for common drivers. See the database integrations for instrumentation guides per database and language. For PostgreSQL query activity from a database receiver, follow the RDS PostgreSQL evidence recipe; it requires the corresponding Last9 API and Database details UI release.

If a service’s database calls appear in its Traces, its traced database identity can appear here. Receiver-discovered PostgreSQL databases can appear without database client spans; this does not create a service-to-database trace relationship.

Understanding the Databases View

Open Discover → Databases to see datastores detected from trace data and supported database receivers. Trace discovery supports technologies such as MySQL, PostgreSQL, MongoDB, Redis and Aerospike when their client spans report db.system. Receiver-discovered database types depend on the receiver integrations available in your workspace.

Each row in the list shows:

ColumnDescription
Database TypeThe datastore technology, from the db.system attribute
HostThe database host, from net.peer.name
ThroughputQuery rate per minute
Latency (P95)95th-percentile query response time
Error RatePercentage of failed database calls
ServicesNumber of services querying this datastore

Sort by any column, or use the left-side filters to narrow by database type or host.

Database Detail View

Click a database to open its detail view. Trace-derived views include Overview, Queries, Services, Infrastructure, Logs, and Traces. For supported PostgreSQL receiver data, Query activity shows Query sample, Query metrics, Explain plan, and Blocking activity records. Estimated plans are labeled Explain plan (estimated). The tab appears when matching receiver data is available and the required Last9 API and UI release is enabled.

Database detail Overview tab

Trace-derived database data powers the Overview and query performance panels. Receiver-discovered PostgreSQL data adds query and blocking evidence in Query activity; it does not supply trace-derived throughput, latency, errors or service relationships on its own.

The Overview tab reports:

  • Query Performance Index: A single health score for the datastore, with summary cards for throughput, average response time, and error rate
  • Performance charts: Query response time (AVG, P50, P90, P95, P99) and throughput over the selected range
  • Query Type Distribution: Per-operation breakdown with throughput, average and P95 latency, and error rate
  • Top 10 Slowest Queries and Top 10 Most Frequent: Each query links to its correlated traces, pre-filtered by span name, db.system, and net.peer.name
  • Top 10 Services by Load: Which services drive the query volume — each links to that service’s Database tab
  • Recent Slow Query Samples: Actual query text from traces, with execution time

From Database to Service and Back

Discover Databases and the per-service Database tab are two views of the same data:

  • Start from a service when the symptom is service latency and you suspect a query.
  • Start from a database when the symptom is datastore-wide — connection saturation, a slow replica, a hot table — and you need to know every service contributing load.

Best Practices

  • Watch P95 per operation, not per database. A healthy average hides one slow UPDATE path. The per-operation breakdown isolates it.
  • Use a trace link when one is available. The linked trace shows the request context around a query. PostgreSQL receiver query records have a trace link only when the application propagates valid trace context.
  • Check fan-in before scaling up. If one database shows load from many services, the Top Services by Load table tells you whether one consumer is responsible before you reach for a bigger instance.

Troubleshooting

  • A database discovered from traces is missing: confirm the querying service emits database client spans — open the service’s Traces and look for spans with a db.system attribute. For PostgreSQL receiver discovery, verify the collector is connected and sending data for the selected database, and that the required Last9 API release is enabled. If the service itself is missing, see Discover Services prerequisites.

Please get in touch with us on Discord or Email if you have any questions.