AWS's November 18, 2026 RDS update speeds storage initialization after point-in-time recovery, read-replica creation and Single-AZ to Multi-AZ conversion. A database can remain available while that work continues. Before declaring a point-in-time restore ready for your application, record storage progress separately from the application's acceptance result.

Start with the operation you actually ran

Record that this was point-in-time recovery, the restored instance identifier, engine, storage type, and observation time. AWS’s storage guide distinguishes these operations from restoring directly from a DB snapshot, which uses background lazy loading. Do not apply the newer initialization behavior to every workflow called a restore.

Read the primary and additional volumes

Use an authorized read-only DescribeDBInstances response. Its top-level StorageOperationStatus and StorageOperationPercentProgress describe the primary volume; where supported and present, inspect the corresponding fields in every AdditionalStorageVolumes entry. AWS documents that volumes can have different operation states. The operation fields disappear when the operation completes, so do not require a permanent stored 100% field.

The volume layout depends on the database engine. Before interpreting absent operation fields, verify that the observation covers the intended current instance and all expected volumes. A partial or stale response is not completion evidence.

A proposed handoff record

Create one row per expected volume, with its identifier, observation time, operation state, reported percentage if present, and the source of the observation. Keep a separate section for an application owner’s approved read-only workload check and its actual outcome. Leave those fields blank until the check has run.

For a hypothetical two-volume recovery, an available instance plus a still-initializing second volume is an open storage observation. A clean primary-volume row must not silently complete the second row. This example illustrates record-keeping only; it is not a measured restore.

Keep completion and acceptance separate

Ask which evidence supports each conclusion: is the restored instance reachable, has the relevant storage operation finished, and did the intended application workload meet its own acceptance criteria? Those observations answer different questions. A faster initialization feature does not establish data correctness or approve a production traffic switch. Keep any unresolved row and its owner visible at handoff.

This checklist describes a proposed review, not an observed restore result.

Primary sources

AWS: fast storage volume initialization announcement

RDS storage guide

RDS point-in-time recovery guide

Conceptual illustration of two storage volumes with example progress indicators, a database checklist, and a separate application review.
Conceptual illustration of two storage volumes with example progress indicators, a database checklist, and a separate application review.