CockroachDB: The Definitive Guide

Chapter 12. Backup and Disaster Recovery

High availability is the only option for most modern applications. In previous chapters, we’ve seen how to configure a distributed CockroachDB cluster that can survive all but the most extreme circumstances.

Nevertheless, even a cluster with the most resilient replication scheme may encounter a circumstance from which it cannot recover. There might be a coordinated failure of multiple data centers—as a result of a cyber attack, for instance—that renders all the data within the cluster irrecoverable.

There are also circumstances in which the cluster infrastructure remains intact, but the data within it becomes corrupted. For instance, a database administrator (DBA) might inadvertently modify data thinking that they are working on the development system, a ransomware attack might attempt to wipe all the data, or an application bug might subtly corrupt data over time.

While CockroachDB has a number of features that can assist you when recovering from data corruption (primarily based on the AS OF SYSTEM TIME clause), backups provide the ultimate insurance policy against loss of or corruption of data in your databases.

Backups also provide some ability to recover the stage of data in the relatively distant past. This might be required for regulatory purposes or in the event that data becomes so logically corrupted that a “do-over” from some point in time is required.

Backup Versus High Availability

In distributed databases such as CockroachDB, ...