When it all goes Wrong (with Postgres)
summary
Practical, diagnostic-first talk that lists actual failure rates and fixes: about 95% application-caused incidents, ~4% autovacuum, ~1% exotic bugs; recommends turning overcommit off (and warns about containers), shows live tools like perf and pg_stat_statements, and walks CPU/memory/disk/backends failure patterns with concrete commands (iostat, iotop, htop, bwm‑ng, gdb) and remediation steps (rate‑limit, add indexes, scale or move reports to followers). Pick it if you want hands‑on triage, skip it for high‑level theory.
AI-written summary · about
chapters (13)
- 0:00Start
- 1:05coming from
- 2:01same sorts of problems
- 2:46the problem with Postgres
- 4:02complicated system
- 4:31using the database (too much)
- 5:44hard to convince
- 8:06snowball
- 9:12assumptions
- 12:49system resources
- 24:08disk
- 25:37tools of the trade: perf
- 33:10flirting with disaster
related talks
36mPostgreSQL worst practices
42mPostgres: stories that will scare you
01h 00mThe 10 Deadly PostgreSQL Sins
47mPostgreSQL worst practices
35mThe PostgreSQL Twin Bill: The 7 Deadly PostgreSQL Mistakes & Review of What's New in PostgreSQL 14
39mDeep dive into PostgreSQL internal statistics
41mApplication Development Challenges with Postgres
51mHow Postgres is misused and abused in the wild