Разгоняем PostgreSQL. Личный опыт. (Перезалив)

Калькулятор параметров PG (PGTune) Параметры из калькулятора: ALTER SYSTEM SET max_connections = ’100’; ALTER SYSTEM SET shared_buffers = ’6GB’; ALTER SYSTEM SET effective_cache_size = ’18GB’; ALTER SYSTEM SET maintenance_work_mem = ’1536MB’; ALTER SYSTEM SET checkpoint_completion_target = ’0.9’; ALTER SYSTEM SET wal_buffers = ’16MB’; ALTER SYSTEM SET default_statistics_target = ’100’; ALTER SYSTEM SET random_page_cost = ’1.1’; ALTER SYSTEM SET effective_io_concurrency = ’200’; ALTER SYSTEM SET work_mem = ’20971kB’; ALTER SYSTEM SET min_wal_size = ’2GB’; ALTER SYSTEM SET max_wal_size = ’8GB’; ALTER SYSTEM SET max_worker_processes = ’6’; ALTER SYSTEM SET max_parallel_workers_per_gather = ’3’; ALTER SYSTEM SET max_parallel_workers = ’6’; ALTER SYSTEM SET max_parallel_maintenance_workers = ’3’; Собственные: ALTER SYSTEM SET parallel_setup_cost = ’500’; ALTER SYSTEM SET cpu_operator_cost = ’’; ALTER SYSTEM SET cpu_tuple_cost = ’’; ALTER SYSTEM SET autovacuum_work_mem = ’1GB’; ALTER SYSTEM SET autovacuum_naptime = ’5s’; ALTER SYSTEM SET autovacuum_vacuum_cost_limit = ’8000’; ALTER SYSTEM SET autovacuum_vacuum_cost_delay = ’2ms’; ALTER SYSTEM SET autovacuum_vacuum_scale_factor = ’0’; ALTER SYSTEM SET autovacuum_analyze_scale_factor = ’0’; После скольки мертвых записей в таблице начинать её автовакуумить (подбираете сами): ALTER SYSTEM SET autovacuum_vacuum_threshold = ’10000’; После скольки мертвых записей в таблице начинать сбор статистики (подбираете сами): ALTER SYSTEM SET autovacuum_analyze_threshold = ’500’; ALTER SYSTEM SET synchronous_commit = ’off’; ALTER SYSTEM SET wal_compression = ’on’; ALTER SYSTEM SET full_page_writes = ’on’; ALTER SYSTEM SET log_autovacuum_min_duration = ’500ms’; Выборка параметров с фильтром: SELECT name, setting FROM pg_settings where name like ’%autovac%’; Включение больших страниц памяти в GRUB: default_hugepagesz=1G hugepagesz=1G hugepages=СвоёЗначение transparent_hugepage=never Рекомендации от Microsoft по тюнингу параметров ядра: Мои параметры ядра: # VM = 0 = 0 = 104857600 = 1073741824 = 300 = 100 =2 =100 = 1 # Kernel = 0 = 0 = 50000000 = 2 = 100000000 = 10000000 # Network = 1 = 1 = 1 = 1073741824 = 1073741824 = 1048576 16777216 1073741824 = 1048576 16777216 1073741824 = 30 = 30 = 20 = 1048576 16770216 1073741824