AWS Adds Do-Over Feature to Aurora Database
Amazon Web Services, which added new capabilities to its Aurora scale-out relational database last year, said this week it is including an “Undo” option with the production platform.
The Amazon Aurora Backtrack feature “is as close as we can come, given present-day technology, to an Undo option for reality,” Jeff Barr, chief evangelist at AWS, explained in announcing the upgrade in a blog posted Thursday (May 10).
The backtrack feature is enabled when launching a new Aurora database clusters. Enabling the feature involves specifying how far back in time to “rewind,” Barr explained. The feature then “will try to retain enough log information to support that window of time” up to 72 hours, according to the backtrack configuration settings page.
Aurora uses a log-structured storage system in which each change in a database generates a new log record. AWS said enabling the backtrack feature provisions a first-in, first-out buffer in the cluster to store the resulting log sequence numbers. “This allows for quick access and recovery times measured in seconds,” Barr said.
In most cases, a database cluster would rewind to a point in time just before an error occurred. The database cluster would then be restored to the previous, stable state without creating a new cluster. Among the use cases are restoring a test database after running tests that make changes in the database, Barr added.
AWS said the Aurora Backtrack feature is available now in all AWS regions supporting Amazon Aurora. Barr noted that backtrack feature applies to newly deployed MySQL-compatible Aurora database clusters and MySQL-compatible clusters restored from a backup. Users must then “opt-in” when creating or restoring a cluster. “You cannot enable it for a running cluster,” Barr noted.
Last fall, the cloud giant (NASDAQ: AMZN) added support for multiple Aurora availability zones for database reads and writes. It also added support in multiple regions for its globally distributed databases.