Infra, Illustrated.
AMAZON RDS · RECOVERY CONSOLE

RDS Backup Retention

See how RDS uses automated snapshots and transaction logs to maintain a moving window for point-in-time recovery.

TEACHING MODEL · orders-dbINTERMEDIATE · VISUAL BRIEFPublished Apr 14, 2026 · Reviewed Sep 15, 2026Amazon RDS DB instances; Aurora has separate backup behavior
ACTIVE VIEWRecovery window
direct flow reused / conceptual
orders-db · 7-day retention
01 / TIME

The recovery window moves.

DAY 8 · NOWSeven days of illustrated recovery
EARLIEST ILLUSTRATED · DAY 2CURRENT EXAMPLE · DAY 8

At Day 9, Day 2 leaves the illustrated window and Day 3 becomes the earliest illustrated day. For a real restore, use the earliest and latest times RDS reports.

02 / RECOVERY PATH

A restore starts a new instance.

01 · BASELINESystem snapshotStored database state
02 · CONTINUETransaction logsMove toward chosen time
03 · RESULTNew DB instanceSource unchanged
SELECTED TIME · DAY 2AVAILABLE IN THE DAY 8 EXAMPLE

Use RDS’s reported range to choose an actual timestamp; this day label is only a teaching marker.

03 / STORAGE LAYER

Only changed blocks are newly stored.

VOLUME STATEDay 1 · full baselineSix written blocks are stored.
RESTORABLE VOLUME
BLOCK AA1Stored now
BLOCK BB1Stored now
BLOCK CC1Stored now
BLOCK DD1Stored now
BLOCK EE1Stored now
BLOCK FF1Stored now
newly stored block references an earlier stored version
SNAPSHOT HISTORYSolid = stored that day · dashed = reused
DAY / NEWABCDEF
Day 16 newA1B1C1D1E1F1
Day 22 newA1B1C2D1E2F1
Day 31 newA1B3C2D1E2F1
Day 41 newA1B3C4D1E2F1
Day 51 newA1B3C4D1E5F1
Day 61 newA1B3C4D1E5F6
Day 71 newA1B3C4D7E5F6
Day 81 newA1B3C4D8E5F6
Day 1 can expire without breaking later restore points.

In this model, C1 and E1 are no longer referenced and can be removed. A1, B1, D1, and F1 remain because later restore points still need them.

These are documented EBS snapshot rules used to illustrate RDS incrementality. RDS manages its real backup dependencies internally; this block graph is not exposed by RDS.

04 / LIFECYCLES

Four names, different clocks.

orders-dbDAY 2 · RECOVERY QUESTION
AUTOMATED BACKUPRolling rangeRDS manages snapshots + logs; Day 2 can age out.
MANUAL SNAPSHOTSaved stateYou create and delete it; one captured state persists.
FINAL SNAPSHOTManual at deletionCreated on request when the source is deleted.
RETAINED AUTOMATEDTemporary setExisting data ages out; nothing new is added.
05 / DELETE DECISIONS

Two choices, three survivors.

SOURCEDelete orders-db
FINAL SNAPSHOTSaved until you delete itOne restorable state
AUTOMATED SETKept temporarilyOld range continues to expire
EXISTING MANUALUnaffectedIndependent of source deletion
06 / SERVICE BOUNDARIES

Check what the setting really controls.

0 DAYSAutomated backups off
1–35 DAYSConfigurable DB instance range
LOG UPLOADEvery five minutesLatest restorable time can still lag now.
INSTANCE STATEAvailable requiredAutomated backups pause in states such as storage_full.
STOPPED INSTANCEClock pausesStopped time does not count toward retention.
RESTORE RESULTNew instanceSource is not rewound in place.
07 / STORAGE DRIVERS

Retention is not a copy count.

CHANGED BLOCKSData churn
MANUAL / FINALSaved snapshots
AFTER DELETIONRetained backups
REGIONAL BACKUP STORAGEMeasure actual usageAllowance and charges depend on running storage and current RDS pricing.

The console shows cost drivers, not an estimate. Check billing metrics and current AWS pricing for your Region and workload.

08 / OPERATING LOOP

Prove the recovery path before you need it.

  1. 01Choose retentionStart from the oldest recovery time your system needs.
  2. 02Observe the rangeRead earliest and latest restorable times from RDS.
  3. 03Preserve a pointTake a manual snapshot when one state must last longer.
  4. 04Decide before deletionFinal snapshot and retained automated backups are separate.
  5. 05Test a restoreCreate a new instance and verify the application can use it.