Start here. This is the direct spoken answer to practice first.
Why this question matters
Disaster recovery is not the same as ordinary high availability. It prepares the product, data, tools, and people to restore acceptable service after a wider failure, with cost matched to business impact.
I would classify the critical user flows and agree on two measurable targets: RTO for how long recovery may take and RPO for how much recent data can be lost. Those targets drive backup frequency, retention, replication, and whether recovery uses restore, cold standby, warm standby, or active-active capacity. The plan would name the systems and data needed for each flow, the restore order, the decision owner, and the validation that proves service is usable. Backups would be encrypted, protected from the primary failure boundary, monitored, and regularly restored into an isolated environment.