Database restored? Verify customer workflows before calling recovery complete
A database restore can finish quickly while customer-facing features still fail, so the two should be recorded separately. The article suggests choosing one critical workflow, such as booking, and defining what a pass looks like, including that a pre-recovery booking is visible and a new booking persists after reload. Restores should go to an isolated environment with test accounts and sandbox integrations so the drill cannot charge cards or notify customers. Teams should log only what was actually verified, along with the restore point and any pre-restore records found. In a fictional drill, the database was queryable after 18 minutes but the booking workflow only passed after 31 minutes, and the drill did not measure incident detection or traffic switching.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.


Discussion (0)
Log in to join the discussion and vote.
Log in