Posts

DTCC 2026 talk: zero-downtime major version upgrade strategies and practices for Ping An RASESQL ultra-large databases

Deep dive into TCP keepalive, FIN vs RST, and connection pool probing mechanisms — with hands-on packet capture tests

A real UUID v4 collision on HackerNews sparked 479 upvotes and 347 comments. Combined with UUID v4 vs v7 performance benchmarks on PG 16, here's how to choose.

The 2026 No.2 Xinchuang database list released: 23 products pass, Ping An, UnionPay, China Mobile, and China Telecom self-developed databases debut. A DBA's interpretation and reflections.

Bruce Momjian's 70-slide talk at PGDay Armenia, from Transformer vector spaces to a pretzel inventory system, dissecting how MCP beats RAG and how far we are from production.

Open-Source Ecosystem Summit & PostgreSQL Summit 2026 — Challenges and strategies for PG memory management under cloud Cgroup constraints

From Jekyll to Hugo, from GitHub Pages to a custom domain, the journey of launching a blog with 76 articles and reflections along the way.

Analyzing the 'pre-existing shared memory block' error when a PostgreSQL instance fails to start after OOM, with a deep dive into PG's three shared memory types and how to locate and clean up residual SysV shared memory

A deep reflection on the essence of DBA operations and the limitations of AI Ops in the AI era, exploring non-technical barriers such as consensus, responsibility, and security, as well as changes in writing and learning

A 2025 summary of PostgreSQL operations experience covering CPU, memory, IO, WAL, autovacuum, locks, partitioned tables, connection pools, statistics, and other core operational topics with practical tips and troubleshooting