PostgreSQL Parallel Query and JIT Cheat Sheet

EXPLAIN (ANALYZE, BUFFERS, VERBOSE)
SELECT tenant_id, sum(total)
FROM app.orders
GROUP BY tenant_id;

Note: Run only against a safe read-only workload; compare Workers Planned and Workers Launched.

BEGIN;
SET LOCAL max_parallel_workers_per_gather = 4;
EXPLAIN (ANALYZE, SETTINGS) SELECT avg(total) FROM app.orders;
ROLLBACK;

Note: SET LOCAL disappears at transaction end; it does not override cluster-wide capacity.

BEGIN;
SET LOCAL parallel_setup_cost = 500;
SET LOCAL parallel_tuple_cost = 0.05;
EXPLAIN (SETTINGS) SELECT sum(total) FROM app.orders;
ROLLBACK;

CREATE OR REPLACE FUNCTION app.net_amount(numeric, numeric)
RETURNS numeric
LANGUAGE sql IMMUTABLE PARALLEL SAFE
RETURN $1 - $2;

Note: Only use PARALLEL SAFE after auditing every operation and called function; function ownership is required to alter it.

SELECT current_setting('transaction_isolation') AS isolation,
       current_setting('max_parallel_workers_per_gather') AS per_gather;

BEGIN;
SET LOCAL jit = on;
SET LOCAL jit_above_cost = 10000;
EXPLAIN (ANALYZE, SETTINGS) SELECT sum(total * tax_rate) FROM app.order_lines;
ROLLBACK;

Note: Look for a JIT section and compare total execution across repeated representative runs.

SELECT name, setting, unit, source
FROM pg_settings
WHERE name IN ('max_worker_processes', 'max_parallel_workers', 'max_parallel_workers_per_gather', 'jit', 'jit_provider')
ORDER BY name;