One of the most sought-after features of a PostgreSQL high availability cluster is the ability to upgrade between major versions while still remaining online. By this, we mean upgrading from PostgreSQL 10 to 11, for instance.
Until the advent of logical replication, such a procedure was a literal impossibility. Even though pg_upgrade was a vast improvement over the previous dump and restore process, it still required a short outage to rebuild the database catalog and copy or link the data files. In a time-sensitive environment, even such a truncated outage of a few minutes can be disruptive and expensive.
A multi-master design changes literally everything in a way that's difficult to truly believe. Because of the use of logical replication, it's possible to mix different PostgreSQL versions within the same cluster. This makes...
