Physical data replication has a small disadvantage—it requires an identical configuration of the servers that are in sync. Simply speaking, the data has to be placed at the same locations on the filesystem. Another thing is that every change in the data files will be replicated, even if that doesn't change the data itself. For example, this happens when the VACUUM command is executed on the master that removes dead tuples from tables, or the CLUSTER command that rearranges rows.
Even if an index is created on a table, it's not the CREATE INDEX command that's sent to the standby, but the contents of the data file with the index data. This can create excessive load on the network and can become a bottleneck for a system ...
