Database

Articles about Database.

Browse by category

There are currently 16 articles in this category. Start with any entry and expand through tags when you want nearby topics.

Why SSDs alone cannot determine PostgreSQL random_page_cost

random_page_cost reflects assumptions about page access, including caching. Understand its relationship to seq_page_cost, physical row order, and tablespace overrides before testing a value such as 1.1.

Database Postgresql

2026-10-18

The Postgres Ecosystem Foundation Proposal: Keeping Extensions Maintained

A PGConf.EU 2026 session will discuss continuity for Postgres extensions and tools. Its questions about succession and shared maintenance costs have practical consequences for dependency management and upgrade planning.

Database Postgresql

2026-10-11

PostgreSQL min_wal_size: recycled capacity versus retained history

min_wal_size sets a floor for recycling WAL files for future writes. It serves a different purpose from retaining history for standbys, so burst-performance tuning and replication continuity need separate decisions.

Database Postgresql

2026-10-04

PostgreSQL parallel scans: why minimum sizes also affect worker counts

PostgreSQL uses minimum parallel scan sizes both to admit scan candidates and to estimate worker counts. PostgreSQL 18 documentation and code show how to separate size thresholds, planner costs, and worker availability when tuning.

Database Postgresql

2026-10-03

AI-ready PostgreSQL: from retrieval to authorized, completed actions

Vibhor Kumar frames AI readiness as introducing AI while retaining control over data, authority, outcomes, and evidence. PostgreSQL and pgvector illustrate why relevance, business constraints, authorization, and external completion need separate controls.

Database Postgresql

2026-09-27

PostgreSQL max_slot_wal_keep_size: balancing WAL retention and recovery cost

max_slot_wal_keep_size limits WAL retention for lagging replication slots. Checkpoint behavior, outage budgeting, and slot monitoring help balance primary storage capacity against consumer recovery costs.

Database Postgresql

2026-09-26

PostgreSQL parallel maintenance: check the limits before adding workers

max_parallel_maintenance_workers caps workers per command. Using PostgreSQL 18 index builds and VACUUM, this guide explains how memory, eligible work, and shared worker capacity constrain parallelism.

Database Postgresql

2026-09-19

PostgreSQL’s 100-argument limit: rethink inputs before raising max_function_args

PostgreSQL’s max_function_args is a read-only value, defaulting to 100; changing it requires rebuilding the server and C extensions. Learn why 51 JSON key/value pairs exceed the limit and how rows or arrays can consolidate arguments.

Database Postgresql

2026-09-13

How far can a Lua-integrated psql go? Inside the VACUUM progress wrapper

An experimental patch embeds Lua in psql so users can define client-side behavior. Using its VACUUM wrapper as the concrete example, this article explains the two-connection monitoring design, the meaning of PostgreSQL progress fields, and the compatibility checks required before testing it.

Database New technology

2026-09-01

Lua Integration Patch Brings Custom Backslash Commands to psql

An experimental patch integrates Lua into psql so users can define their own backslash commands. Using a custom `\my.dt` command that sorts tables by size, this article examines the new extension model and the practical choices available today.

Database New technology

2026-08-27

Choosing UUIDv7 for Primary Keys in PostgreSQL 18

PostgreSQL 18 includes uuidv7() for generating time-ordered UUIDs. This article explains how UUIDv7 changes B-tree behavior compared with UUIDv4, the tradeoff of exposing time information, and a measured path for adopting it in existing tables.

Database Postgresql

2026-08-27

Isolating PostgreSQL JIT Failures with Two GUCs

A PostgreSQL diagnostic technique uses jit_expressions and jit_tuple_deforming to narrow suspected JIT crashes or wrong results to expression compilation or tuple deforming. This article explains their asymmetric relationship and a controlled reproduction procedure.

Database Postgresql

2026-08-17

Why HA Does Not Make a Long PostgreSQL checkpoint_timeout Safe

A longer PostgreSQL checkpoint_timeout can increase WAL replay exposure not only on the primary but also when a standby restarts. This article explains why HA is not a substitute for recovery-time control and connects checkpoints, restartpoints, monitoring, and failure drills into one availability decision.

Database Postgresql

2026-08-16

How PostgreSQL ignore_system_indexes Enables Recovery from Catalog Index Corruption

PostgreSQL's `ignore_system_indexes` setting removes damaged system catalog indexes from the read path so an operator can reach `REINDEX SYSTEM`. This article connects `-P`, single-user recovery, version differences, and post-repair verification.

Database Postgresql

2026-08-16

Why PostgreSQL COUNT(DISTINCT) cannot use parallel aggregation—and how to rewrite it

PostgreSQL cannot use parallel aggregation when an aggregate call contains DISTINCT. Using the primary source’s execution plans and benchmark as the starting point, this article explains a GROUP BY rewrite, NULL semantics, and what to validate with EXPLAIN.

Database Postgresql

2026-08-06

Reading PostgreSQL platform trust through the CALM Platform Test

Vibhor Kumar wrote "The CALM Platform Test" to frame enterprise platform readiness through four dimensions: Changeability, Assurance, Leverage, and Measurability. Read with PostgreSQL logical replication, RLS, pgAudit, pgvector, and pg_stat_statements in mind, it becomes a practical prompt to verify changeability and evidence before placing AI workloads on a data platform.

Database Postgresql

2026-07-29