Query Metrics

The Query Metrics tab groups executions of the same query together, so you can rank queries by how often they run and how long they take. PostgreSQL queries are grouped by normalized query text per instance. MongoDB operations are grouped by query signature per host.

Open it from APM → Database → Query Metrics, and select the engine in the Overview list.

Charts#

Two charts at the top show trends for the queries in the table. Choose the metric for each chart from its dropdown:

• Query count
• Total Duration
• Avg Duration
• P95 Duration

Each line is one query. Hover over a chart to see the value for each query. Your chart choices are remembered.

Query Table#

Table controls:

• Full Query Text: shows complete query text instead of truncating it.
• Customize: choose which columns to show, then click Apply Changes.
• Sort by clicking a column header. Use the search bar at the top of the page to filter by query text.

Query Details#

Click a query to open its detail panel.

• View Query Samples opens the Samples tab filtered to this query.
• Filter to Query adds the instance and query to the page search, so every tab shows only this query.

The top of the panel shows the instance, the Query text, Tags collected from recent executions, and Avg Duration, Total Duration, and Count charts.

PostgreSQL#

Click the arrow next to Query to format the SQL. Tags include db.instance, db.namespace (database), service.instance.id, backend type, user.name, and the client address.

MongoDB#

The Query section has five views of the latest execution: Statement, RawQuery, Cursor, Locks, and FlowControl. Tags include the host, namespace, operation, plan summary, and query ID.

Explain Plans (PostgreSQL) #

The Explain Plans tab shows the most recent plan the agent captured for the query within the selected time range. Plans come from Top Query Collection and are produced with EXPLAIN, so the query is never executed to get them.

Two views are available:

• Explained: a readable plan tree.
• Json: the raw EXPLAIN (FORMAT JSON) output.

The Explained view contains:

• Plan summary: the root operation (for example Seq Scan, Hash Join, Index Scan), Planner cost, Est. rows (or Actual rows when present), and Planning and Execution time when available.
• Sequential scan detected: a warning shown when any step reads a full table. Consider adding an index or a more selective filter if the query is slow.
• Tables, Indexes, and Access: the tables and indexes the plan touches, and the scan and join methods it uses.
• Plan steps: every plan node, nested under its parent. Each node shows its type, Planner cost (startup → total), estimated rows, loops, workers, the table or index it reads, and details such as Filter, Index Cond, Hash Cond, and Sort Key.

If no plan is available, the tab shows "No explain plan found for this query in the selected time range." See PostgreSQL troubleshooting.

Wait Events and Blocking (PostgreSQL) #

Every PostgreSQL query sample records the wait event the session was in when sampled. Wait events show why a query is slow: waiting on locks, disk, the client, or internal resources.

Wait event groups appear in the Wait Events column of the query table, in the Load By Wait Group column of the instance list, in the instance Average Load chart, and in the Samples tab.

Blocking: when a session waits on a Lock, PostgreSQL reports which sessions block it. Database Monitoring uses this to separate root blocking queries (the queries holding the lock) from waiting queries. The Blocking Overview table lists:

Blocking Samples lists the individual samples, including Blocking Pids. You can find blocking data in the instance Summary → Blocking Queries section, the instance Queries tab, and each query's Blocking Activity tab.