Once streaming replication has been set up, it works flawlessly without too much administrator intervention. However, in some cases it might make sense to halt replication and to resume it at some later point. Why would anybody want to do that?
Consider the following use case: you are in charge of a master/slave setup, which is running some rubbish CMS or some dubious forum software. Suppose you want to update your application from awful CMS 1.0 to dreadful CMS 2.0. Some changes will be executed in your database, which will instantly be replicated to the slave database. What if the upgrade process does something wrong? The error will instantly be replicated to both nodes due to streaming.
To avoid instant ...
