Unconference - Stream 3 - 13:15-14:15 - PGCon 2022

Improve PostgreSQL's logical replication with advanced filtering, synchronous commit settings, and remote apply feedback for enhanced performance and consistency.

Key takeaways

The current feedback mechanism in PostgreSQL’s logical replication can be weak and doesn’t provide immediate feedback, leading to inconsistent database states.
Introducing an option to filter out transactions from specific origins can improve logical replication performance and latency.
The concept of “standby” can be confusing, even in the documentation, and using a more precise term like “physical standby” or “logical standby” can help clarify the concept.
There are different synchronous commit settings in PostgreSQL, and setting local as the default can cause issues with performance.
Using remote apply in PostgreSQL can provide query consistency, but it requires careful configuration and monitoring.
The origin ID is used to identify the replication origin, and it’s possible to filter out transactions from specific origins based on this ID.
The feature for remote replay feedback in PostgreSQL’s logical replication is not well-documented and can be improved for better performance and consistency.
The synchronous commit setting in the subscriber can be adjusted to improve performance, but it requires careful configuration and monitoring.
The default value for synchronous commit is off, but setting it to local can cause issues with performance.
The replication protocol in PostgreSQL’s logical replication is based on a feedback mechanism, which can be improved for better performance and consistency.

Unconference - Stream 3 - 13:15-14:15 - PGCon 2022

More talks