CROSS-REGION SOFTWARE FAILURE RECOVERY

Techniques are disclosed relating to software failover across multiple computing regions. Software entities in a first computing region implement failure recovery routines on an as-needed basis, which can include utilizing code in a recovery API. These software entities can subscribe to receive failure notifications for that region. When an indication of a regional failure is received (e.g., indicating that some specified threshold of servers in the first computing region have failed or are likely to fail), a failure notification can be published. The computing region can then use a list of subscribing software entities to initiate recovery routines for those entities to failover to a different one of the computing regions. In some implementations, software entities can handle failover for dependent components. Thus, when failover is initiated for a particular software product, the product can initiate failover for any constituent components (e.g., services that make up the software product).