How to interpret pg_stat_statements

pg_stat_statements is a supplied PostgreSQL extension that tracks planning and execution statistics for normalized statement fingerprints. Similar statements can be combined when their parsed structures match after literal constants are normalized.

Each row represents a combination that includes database, user, query identifier, and whether the statement is top-level. The counters are cumulative, not a live list of currently running queries.

Extensions can be installed outside public. This read-only catalog query identifies the schema that owns the extension objects in the current database.

SELECT
  e.extname AS extension_name,
  n.nspname AS extension_schema,
  e.extversion AS extension_version
FROM pg_extension AS e
JOIN pg_namespace AS n ON n.oid = e.extnamespace
WHERE e.extname = 'pg_stat_statements';

No row means the extension is not installed in this database, even if the module is active for the PostgreSQL server. A relation-not-found error can also mean the extension schema is absent from the role's search path; use the discovered schema when qualifying the view.

This bounded query deliberately omits the query column. Run it with a role that has the required statistics visibility and with the extension schema available on the search path.

SELECT
  queryid::text AS queryid,
  calls,
  total_exec_time,
  mean_exec_time,
  rows,
  shared_blks_hit,
  shared_blks_read,
  temp_blks_written,
  wal_bytes
FROM pg_stat_statements
WHERE dbid = (
  SELECT oid
  FROM pg_database
  WHERE datname = current_database()
)
ORDER BY total_exec_time DESC
LIMIT 50;

The optional pg_stat_statements_info view reports when all statement statistics were last reset and how often entries were deallocated after the configured entry limit was exceeded. Locate it in the extension schema rather than assuming public.pg_stat_statements_info.

Without a reset timestamp, label the window unknown. Restarts, explicit resets, configuration changes, and entry deallocation can all limit comparisons. A fingerprint's queryid also has limited stability guarantees and should not be assumed stable across PostgreSQL major versions.

• The top cumulative contributor is not necessarily the slowest individual call or the current bottleneck.
• Two visually different statements can share a normalized entry, and identical-looking text can produce different identifiers when its parsed meaning differs.
• A cache-hit percentage is not a direct storage, disk-latency, or provider-cost measurement.
• One snapshot cannot show whether a contributor is accelerating; compare compatible counter differences across a known interval.
• Resetting statistics to simplify analysis destroys the existing observation window and should not be a routine diagnostic shortcut.

The pg_stat_statements analyzer ranks bounded fingerprints by total contribution, frequency, average execution time, planning time, temporary blocks, and WAL counters. Manual CSV, TSV, and JSON analysis stays in the browser; query-text columns are discarded before analysis and export.

An eligible connected database can return the same bounded counters without copy and paste. DatabaseMeter does not retain SQL text, does not reset PostgreSQL statistics, and does not turn a cumulative ranking into an automatic remediation.