Menu

Post image 1
Post image 2
1 / 2
0

War Story: How a PostgreSQL 17 Deadlock Caused 30 Minutes of Downtime for Our E-Commerce Checkout Flow

DEV Community·ANKUSH CHOUDHARY JOHAL·about 1 month ago
#4tH4wa2Z
Reading 0:00
15s threshold

At 14:17 UTC on October 12, 2024, our e-commerce checkout flow ground to a halt for 30 minutes, costing $142,000 in lost sales and 12,000 abandoned carts. The root cause? A silent, self-deadlock in PostgreSQL 17’s new parallel query executor that no monitoring tool caught until we dug into raw WAL logs. 📡 Hacker News Top Stories Right Now Credit cards are vulnerable to brute force attacks (60 points) Ti-84 Evo (80 points) New research suggests people can communicate and practice skills while dreaming (120 points) Show HN: Destiny – Claude Code's fortune Teller skill (24 points) Ask HN: Who is hiring? (May 2026) (193 points) Key Insights PostgreSQL 17’s parallel sequential scan (PSS) feature introduces a 0.4% deadlock risk for transactions with overlapping UPDATE/SELECT FOR UPDATE patterns on partitioned tables, up from 0.01% in PG15. pg_stat_statements 1.10 (bundled with PG17) adds deadlock candidate tracking, reducing root cause analysis time from 4 hours to 12 minutes.…

Continue reading — create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.

Read More