Database Operations
Database Operations summarizes database spans by system, namespace, collection, operation, and statement. Use it to find the query or datastore operation consuming time or contributing errors.
The tab appears only when enabled for a deployment with the database rollup configured.
Overview Charts
Four charts follow the current environment, service, search, filter, and time scope:
| Chart | Meaning |
|---|---|
| Database operations | Call volume |
| Database latency P95 | P95 execution duration |
| Database error rate | Percentage of failed operations |
| DB time (ms per second) | Aggregate time spent in database calls per second |
Drag across a chart to narrow the whole application to that interval. Configured annotations appear on these charts.
Search and Filters
Topbar search matches database system, collection, operation, and statement. The filter sidebar provides facets for:
- System.
- Namespace.
- Collection.
- Operation.
- Raw span kind.
The service and environment selectors remain in the topbar.
Operations Table
| Column | Meaning |
|---|---|
| System | Database technology, such as PostgreSQL or Redis |
| Namespace | Database or logical namespace |
| Collection | Table, collection, or equivalent entity |
| Operation | Database operation name |
| Statement | Captured statement or span name |
| Calls | Number of executions |
| P95 | P95 execution duration |
| Errors | Failed execution count |
Click a heading to sort server-side results. Use Load more in scrolling mode or page controls in pagination mode. Select a row to open its resizable detail drawer.
Database Operation Detail
The drawer contains:
- A summary with service, environment, span name, database dimensions, calls, P95, errors, and total database time.
- A syntax-highlighted statement.
- A per-minute Database time chart.
- The Top 10 slowest executions from raw spans.
The chart remains rollup-backed. Slowest executions are loaded separately, so a raw-query failure does not remove the operation summary or chart.
Raw Execution Window
The slowest-executions section displays the exact absolute range queried. The start is moved forward when necessary so APM never reads data older than the administrator's raw trace limit. No progressive one-hour search applies here; the top ten are selected over the entire allowed portion of the chosen range.
Select an execution to open its exact trace and database span in traceX.
Use Cases
Find the Query Behind a Slow Endpoint
- Scope to the service that owns the endpoint.
- Sort the operations table by P95 to find its slowest statement.
- Open the drawer, read the statement, and select one of the ten slowest executions to see it inside a full trace.
Cut Total Database Time
- Read the DB time chart rather than P95: a fast query called constantly can cost more than a slow one called rarely.
- Sort by calls to find the chatty operations.
- Check whether the caller can batch or cache them.
Chase Database Errors
- Filter the sidebar to the system or collection reporting failures.
- Sort by errors and open the failing operation.
- Follow an execution into traceX to see what the application did next.
FAQ
Should I chase P95 or DB time first?
Start with DB time. P95 finds the slowest single execution, while DB time counts duration multiplied by frequency, so it surfaces the cheap query called thousands of times that actually dominates the request. The two often point at different operations.
Why does the slowest-executions list cover less time than the chart above it?
The chart reads the rollup and the list reads raw spans, which have their own retention limit. The list states the absolute interval it queried, and its start is moved forward when the selection reaches past that limit. There is no progressive one-hour search here; the top ten are chosen over the whole allowed interval.
Why is the Database Operations tab missing?
The tab appears only for deployments where an administrator has enabled it and configured the database rollup. Database spans alone are not enough.
Related Guides
- Getting Started - Scope, filters, and trace limits
- Services - The service issuing the query
- Traces - The request around a database span
- pgX - PostgreSQL monitoring from the database side