INFRASTRUCTURE LEARNINGINFRASTRUCTURE LEARNING
インフラ大全インフラ大全IT INFRASTRUCTURE / LEARNING
Infrastructure learning guide
Find a lesson or lab/

Esc to close · Search terms stay on your device

LEARNING日本語

Topic lessonTopic: S3: versions, retention, replication and cost

S3 Replication explained: CRR, SRR and what gets copied

S3 Replication asynchronously copies eligible objects to another bucket. CRR crosses Regions; SRR stays within one Region. Saving a configuration is different from confirming that the required data has been replicated.

What you will learn

Distinguish new and existing targets, versioning, permissions, encryption, deletes and Lifecycle; explain how to verify replication results.

Prerequisite

Know Versioning, ACLs/policies and Lifecycle. The synthetic example is one-way replication from general purpose bucket A to B, both versioning-enabled. Its report is written after configuration and matches the rule. No AWS connection is made.

Match the source, selection rule and destination

One-way live replication: verify completion before assuming the copy exists
One-way live replication: verify completion before assuming the copy exists1: Source bucket A — Select eligible versions written after configuration, not every earlier object. — Replication rule. 2: Replication rule — Copy when permissions and other conditions are met. Reverse replication and application failover are separate. — Destination B12Source bucket AReplication ruleDestination BOne-way live replication: verify completion before assuming the copy exists1: Source bucket A — Select eligible versions written after configuration, not every earlier object. — Replication rule. 2: Replication rule — Copy when permissions and other conditions are met. Reverse replication and application failover are separate. — Destination B12Source bucket AReplicationruleDestination B
  1. Source bucket A → Replication ruleSelect eligible versions written after configuration, not every earlier object.
  2. Replication rule → Destination BCopy when permissions and other conditions are met. Reverse replication and application failover are separate.
Source bucket A
Versioning enabled; a new eligible report version is stored.
Replication rule
Selects objects, destination and the role used by S3.
Destination B
Versioning enabled; copying is asynchronous. Check completion and read access.
Region placement and object age are separate choices
NameDistinction
CRRLive replication between buckets in different Regions
SRRLive replication between buckets in the same Region
Batch ReplicationAn on-demand job for selected existing objects and other eligible targets

Choose an option to see the result and explanation.

A live replication rule is created today. Does it automatically copy all objects stored last month?
Read the answer and reason

Answer: No; consider existing objects separately for Batch Replication · Live replication targets eligible later writes. Existing data needs its own selection, job, permissions and cost review.

Separate the configurator’s permissions from the role S3 uses

The identity configuring replication differs from the role assumed by S3 to perform it. Review trust, source-version read permissions and destination replication permissions. Cross-account destinations also need permissions granted from their side.

Review prerequisites for the relevant object version
ItemWhat to review
VersioningEnabled on both buckets; suspension at the destination causes failures
SelectionMatch prefix/tags and distinguish pre-existing objects
OwnershipReview source read access, destination ownership and associated permissions
ProtectionA source using Object Lock imposes destination requirements; review retention

For cross-account copies, decide whether the destination owns the replica or source ownership is retained. Object Ownership settings such as Bucket owner enforced and owner override affect that choice. Arrival alone does not establish that the destination application can read it.

For KMS-encrypted versions, check explicit selection and key access

AWS Key Management Service (KMS) manages encryption keys. SSE-KMS uses KMS for server-side encryption; DSSE-KMS adds dual layers. Replicating these objects requires explicit opt-in. Review source decryption, destination encryption, key policies and Regions. S3 read permission alone does not establish KMS access.

Choose an option to see the result and explanation.

An SSE-KMS replication configuration was saved. Does that alone prove the data was replicated?
Read the answer and reason

Answer: No; verify key conditions and per-version replication results · Configuration acceptance differs from data processing. Verify that source and destination key permissions work for the actual replication result.

Different delete operations affect the destination differently

Live replication using the current Filter-based configuration
Source changeDestination behavior
DELETE without version IDDelete markers are not replicated by default; supported rules can explicitly enable this
Permanent deletion of a specific versionThat deletion is not replicated to the destination
Delete marker created by LifecycleNot replicated, even with delete marker replication enabled
Change to bucket Lifecycle configurationThe configuration itself is not copied to the destination

Delete marker replication does not support tag-based rules. Older V1 configurations behave differently, so inspect existing configurations before applying this table. A current replicated marker can hide an object from ordinary reads at the destination without removing its retained data versions.

Choose an option to see the result and explanation.

A source version is permanently deleted by specifying versionId=vA. Does replication necessarily delete the destination version?
Read the answer and reason

Answer: No; the version-specific deletion is not replicated · Version-specific deletion, markers and Lifecycle are different. This does not prevent deletion by the destination’s own permissions or Lifecycle rules.

Match PENDING, COMPLETED and FAILED to the required version

Statuses for a basic one-way replication example
StatusInterpretation
PENDINGNot complete; a delay is not automatically a terminal failure
COMPLETEDThe version completed replication; with several destinations review completion for all
FAILEDInvestigate permissions/keys; fixing configuration alone does not resend every failed version
REPLICAIdentifies a replica; does not establish application failover

For FAILED versions, fix the cause and consider reprocessing such as Batch Replication. PENDING/FAILED also affects source Lifecycle actions. Destination Lifecycle uses the original object creation time, not the replica’s arrival time, so review its retention independently.

  1. Define required data and recovery needs; separate Regions, accounts and existing objects.
  2. Review versioning, selection, S3 role, ownership and encryption on both sides.
  3. Verify version status and destinations; record causes and reprocessing targets.
  4. Design destination retention, deletion permissions, application reads/failover and costs separately.

Check, then explain in your own words

  1. Separate CRR/SRR and live/Batch.
  2. Review selection, versioning, permissions and encryption.
  3. Distinguish delete types and destination retention.
  4. Verify per-version completion and reprocessing targets.
Check your understanding: Does enabling replication automatically cover all existing data, deletes and recovery?

No. Review eligibility, delete behavior, permissions, retention and completion separately.

Replication is asynchronous copying of selected data. Reading and recovering the required version depend on destination design and verification too.

Sources and conditions checked

The official references used in this lesson are collected here with their scope, conditions and corresponding sections. Independent expert review and live AWS comparison have not been performed.