Home › Databases

ERROR: canceling statement due to statement timeout (PostgreSQL) — how to fix

ERROR:  canceling statement due to statement timeout
SQL state: 57014

Quick fix: Find where the timeout is set (SHOW statement_timeout;) and either make the query faster (check EXPLAIN (ANALYZE, BUFFERS), add indexes, avoid lock waits) or raise the limit just for that work: SET LOCAL statement_timeout = '5min'; inside the transaction.

What it means

statement_timeout is a per-statement time limit (default 0 = off). When a statement exceeds it, Postgres cancels it and the surrounding transaction is aborted. Someone set it: in postgresql.conf, on the role/database, on the connection string, or in your framework/ORM.

Similar messages with different causes:

Find where the timeout comes from

SHOW statement_timeout;

SELECT name, setting, source, sourcefile
FROM pg_settings WHERE name = 'statement_timeout';

-- role/database level overrides
SELECT coalesce(r.rolname, 'ALL') AS role, coalesce(d.datname, 'ALL') AS db, s.setconfig
FROM pg_db_role_setting s
LEFT JOIN pg_roles r ON r.oid = s.setrole
LEFT JOIN pg_database d ON d.oid = s.setdatabase;

Managed services (RDS parameter groups, Supabase, Heroku) and poolers often set it too. Supabase, for example, sets short timeouts on the anon and authenticated roles.

Cause 1 / Fix 1: the query is slow

Run it with timing data (on a copy or with the timeout raised):

SET statement_timeout = 0;
EXPLAIN (ANALYZE, BUFFERS) SELECT ...;

Look for Seq Scan on large tables with a selective filter and add the missing index:

CREATE INDEX CONCURRENTLY idx_orders_customer ON orders (customer_id);
ANALYZE orders;

Other common fixes: keyset pagination instead of huge OFFSET, avoiding functions on indexed columns in WHERE (lower(email) needs an expression index), and running VACUUM ANALYZE when statistics are stale.

Cause 2 / Fix 2: waiting on locks

The statement may be fast but blocked by another transaction (time spent waiting counts toward statement_timeout). Check:

SELECT pid, pg_blocking_pids(pid) AS blocked_by, now() - query_start AS waiting, query
FROM pg_stat_activity
WHERE cardinality(pg_blocking_pids(pid)) > 0;

Often the blocker is an idle in transaction session. Fix the app so it commits, and set idle_in_transaction_session_timeout to kill stragglers.

Cause 3 / Fix 3: legitimate long work (migrations, reports, backfills)

Raise the limit only for that transaction instead of globally:

BEGIN;
SET LOCAL statement_timeout = '30min';
UPDATE big_table SET ...;
COMMIT;

Or for one role: ALTER ROLE reporting SET statement_timeout = '10min'; (applies to new sessions). For bulk updates, batch them (WHERE id BETWEEN ...) so each statement stays short.

Note: with PgBouncer in transaction mode, a plain SET may leak to other clients or be lost; use SET LOCAL inside a transaction, or set it on the role.

Cause 4 / Fix 4: client-side timeouts

Some drivers send a cancel themselves, which shows up as canceling statement due to user request, e.g. JDBC setQueryTimeout, Django OPTIONS: {"options": "-c statement_timeout=5000"}, or Node pg's statement_timeout / query_timeout config. Check your connection settings before changing the server.

After any of these errors inside a transaction, you must ROLLBACK before the connection can be used again (otherwise: current transaction is aborted).

Last updated 2026-11-10.