Parameters
Resource Usage
44 PostgreSQL 18.6 parameters. Each one has its own animation and explanation.
Background Writer
bgwriter_delay: Sleep between background writer rounds (default 200 ms).bgwriter_flush_after: After this much bgwriter output, ask the kernel to start writing it to disk.bgwriter_lru_maxpages: Max buffers the background writer writes per round (0 disables it).bgwriter_lru_multiplier: How far ahead of predicted demand the background writer cleans (default 2×).
Disk
file_copy_method: How CREATE DATABASE … FILE_COPY copies files: plain copy or filesystem clone.file_extend_method: How data files grow during bulk loads: posix_fallocate or writing zeros.max_notify_queue_pages: Max size of the LISTEN/NOTIFY queue on disk (in 8 kB pages, default = 8 GB).temp_file_limit: Max disk space one process may use for temporary files at once (-1 = unlimited).
I/O
backend_flush_after: After this much data written by one backend, force writeback (0 = off).effective_io_concurrency: How many reads one session keeps in flight ahead of time (default 16 in PG18).io_combine_limit: Largest single read built by merging neighbouring blocks (default 128 kB).io_max_combine_limit: Server-wide ceiling for io_combine_limit (needs a restart).io_max_concurrency: Max asynchronous I/Os one process can have in flight (-1 = automatic, up to 64).io_method: How asynchronous I/O is executed: worker processes, io_uring, or synchronously.io_workers: Number of I/O worker processes used when io_method = worker (default 3).maintenance_io_concurrency: Reads in flight for maintenance work such as VACUUM and ANALYZE (default 16).
Kernel Resources
max_files_per_process: How many files each server process keeps open at once before closing and reopening.
Memory
autovacuum_work_mem: Memory for each autovacuum worker's dead-row list (-1 = use maintenance_work_mem).commit_timestamp_buffers: Cache size for pg_commit_ts (commit time per transaction). 0 = automatic.dynamic_shared_memory_type: How extra shared memory for parallel query is created (posix /dev/shm by default).hash_mem_multiplier: Hash tables may use work_mem × this value (default 2).huge_page_size: Which huge page size to request (0 = the system default, usually 2 MB).huge_pages: Whether the main shared memory segment uses large (2 MB / 1 GB) OS pages.logical_decoding_work_mem: Memory a logical decoding walsender uses per slot before spilling or streaming.maintenance_work_mem: Memory for VACUUM dead-row lists, CREATE INDEX sorts, ADD FOREIGN KEY.max_prepared_transactions: How many two-phase-commit (PREPARE TRANSACTION) transactions may exist; 0 disables it.max_stack_depth: Safety limit on recursion depth (deep expressions, recursive functions).min_dynamic_shared_memory: Shared memory reserved at startup for parallel query, instead of asking the OS each time.multixact_member_buffers: Cache size for pg_multixact/members (who shares a row lock).multixact_offset_buffers: Cache size for pg_multixact/offsets (where each MultiXact's member list starts).notify_buffers: Cache size for the LISTEN/NOTIFY queue (pg_notify).serializable_buffers: Cache size for pg_serial (SERIALIZABLE transaction conflict tracking).shared_buffers: The main shared page cache: every table and index page is read through it.shared_memory_type: How the main shared memory segment is allocated (mmap or System V).subtransaction_buffers: Cache size for pg_subtrans (subtransaction → parent map). 0 = automatic.temp_buffers: Per-session cache for temporary tables (CREATE TEMP TABLE).transaction_buffers: Cache size for pg_xact (commit status of every transaction). 0 = automatic.vacuum_buffer_usage_limit: Size of the ring buffer VACUUM and ANALYZE recycle instead of flooding shared_buffers.work_mem: Memory one sort or hash step may use before spilling to temporary files.
Worker Processes
max_parallel_maintenance_workers: Max workers per CREATE INDEX or VACUUM (default 2).max_parallel_workers: How many of the worker slots parallel query and maintenance may use at once.max_parallel_workers_per_gather: Max workers one Gather node may use; 0 disables parallel query.max_worker_processes: Total background-worker slots: parallel workers, logical replication, extensions.parallel_leader_participation: Whether the session running a parallel query also scans data itself.