Point-in-time recovery
We covered scalability and one way to scale horizontally and failover. We failed over to a node, which contains all transactions, which were committed till the primary went down.
Sometimes, we want to go back in time, retrieve the database cluster at a specific point in time, and then start afresh. For example, we made a release in production last evening. Sanity testing told us things looked fine and we made the system open to users. Sometime this morning, customer support tells us that there is some issue with the system and many orders entered after the release did not look right. Once we are sure that there is a critical bug, the first thing we want to do is to roll back the changes.
On the application side, it's relatively easy. Get the previous release from version control system and deploy it. What about the database? If it's a small cluster (a few GBs), it's always possible to take a full backup before each release and restore from that. In case...
