Citus Con CFP is closed. Thanks to all submitters 🙏 Remember to save the date for Citus Con: An Event for Postgres 2023.
Reducing replication lag with IO concurrency in Postgres 15
PostgreSQL 15 improves crash recovery and physical replication performance of some large and very busy databases by trying to minimise I/O stalls. A standby server might now have an easier time keeping up with the primary.
How? The change in PostgreSQL15 is that recovery now uses the
maintenance_io_concurrency setting (default is 10, but you can increase it) to decide how many concurrent I/Os to try to initiate, rather than doing random read I/Os one at a time. With big and busy databases, when I/O concurrency increases, replication lag can be reduced.