What the heck is a checkpoint, and why should I care?
summary
Explains exactly what a checkpoint writes and why: it flushes dirty pages to disk, records a begin/end checkpoint LSN in the boot page, and relies on write ahead logging so committed transactions survive crashes while uncommitted ones get undone. Shows that checkpoints do not evict pages, how lazy and eager writers differ, and how recovery interval / indirect checkpoint change frequency and IO shape. Includes demos and uses traceflags (3502/3504) to show timing, page counts and millisecond durations.
AI-written summary · about
chapters (5)
- 0:00Start
- 12:57Write Ahead
- 36:53Automatic "Recovery Interval" default zero • What does this value mean? How reliable is this setting? • When do you change default? Changing to over 60 min require "OVERRIDE"
- 38:26Manual Transact-SQL CHECKPOINT 'checkpoint_duration' seconds to complete • Not guaranteed • Advanced option When do you use a Manual Checkpoint? When do you use 'checkpoint duration'?
- 42:21Checkpoint tempdb
related talks
51mWhat the heck is a checkpoint, and why should I care?- 51mWhat the heck is a checkpoint, and why should I care?
44mUnderstanding Write-Ahead Logging (WAL) in PostgreSQL
48mA Deep Dive Into Database Internals
34mThe Computer Wants to Lose Your Data
57mDurability in Postgres
31mPostgreSQL WAL and Commit Synchronization, Optimization Opportunities
26m“Hit me, baby, just one time” – Building End-to-End Exactly-Once Apps w/ Flink