checkpoint_completion_target

checkpoint_completion_target controls how gradually a checkpoint's dirty-page writes are spread across the time until the next checkpoint. A value near 1.0 dribbles the writes out slowly, avoiding a sharp I/O spike at checkpoint time.

Modern PostgreSQL defaults to 0.9, which is what you want in almost all cases — it spreads writes over 90% of the interval. There's rarely a reason to lower it. (Older guides recommended raising it from 0.5; that's already done by default now.)

• checkpoint_timeout
• max_wal_size
• bgwriter_lru_maxpages