Lower binlog_retention_period gradually in small decrements during a low-traffic window instead of dropping it all at once. The docs use 600 as an example value. After each step, watch write latency and the purge progress before going lower. Never slash retention on a busy primary in one change.

Context: Official docs (terraform provider mysql resource): documents the gotcha that drastically lowering binlog_retention_period on Aiven MySQL makes a large batch of binary logs eligible for purge at once, which can stall the MySQL commit path and block writes until the purge completes. Agents tuning retention to save disk can accidentally freeze writes.