Rebuild the DynamoDB settings that a point-in-time restore does not carry over

A newly restored table needs more than recovered items before it is ready for application traffic.

Source facts

DynamoDB point-in-time recovery restores to a new table while the source remains available. AWS lists operational settings that must be set up again after restoration, including auto scaling policies, IAM policies, CloudWatch alarms, tags, Streams, Time to Live and PITR settings. Restore duration depends on data volume and rebuilt secondary indexes; AWS does not guarantee a fixed completion time.

Source: AWS — Restoring a DynamoDB table to a point in time.

Applicability

This concerns the documented DynamoDB point-in-time restore workflow. Restored data and indexes are not evidence that application routing, event processing or retention behavior is ready.

DSE recommendation

DSE recommends preparing a post-restore configuration checklist before an incident. Assign owners for table protection, identity permissions, monitoring, TTL and stream consumers. Decide which indexes are essential before choosing a faster restore that omits them. Keep the restored table isolated until those choices are verified, then approve the application’s transition to the new table and the handling of writes made during recovery. Prepare a comparison sheet with a named owner for every setting that must be restored separately. Make application cutover conditional on closing those gaps and record any intentionally omitted index as an explicit performance or functionality decision for the recovery target.

Verification

Proposed verification: restore an authorized test table and compare each omitted setting against the checklist. Exercise a representative application read and write, the intended stream integration and the selected retention behavior. Record actual recovery elapsed time and remaining gaps instead of converting a completed restore job into an unsupported recovery-time guarantee.

Official reference

AWS — Restoring a DynamoDB table to a point in time. Reviewed October 20, 2026.