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: RDS and SQL Server: management, connections and deployment

Migrate SQL Server to AWS: choose a method and plan cutover

Moving SQL Server to AWS requires more than a database copy: verify compatibility, ongoing changes and where applications write. Use one RDS migration example to compare methods and write-pause budgets, decide when to cut over and understand what returning to the source would require.

What you will learn

Match migration requirements to methods and explain cutover and fallback using final changes, validation and write ownership.

Prerequisite

For readers familiar with SQL Server structure, access, backup/restore and ordinary RDS responsibilities. This is a fictional plan with no connection to AWS or a live database.

Check destination compatibility before copying

Separate copying, final validation and changing the writer
Separate copying, final validation and changing the writer1: Copy data — Copy completion alone is insufficient — Reconcile final changes. 2: Reconcile final changes — Only after required conditions pass — Change the writer12Copy dataReconcile finalchangesChange the writerSeparate copying, final validation and changing the writer1: Copy data — Copy completion alone is insufficient — Reconcile final changes. 2: Reconcile final changes — Only after required conditions pass — Change the writer12Copy dataReconcile finalchangesChange thewriter
  1. Copy data → Reconcile final changesCopy completion alone is insufficient
  2. Reconcile final changes → Change the writerOnly after required conditions pass
Copy data
Choose the initial copy and treatment of ongoing changes.
Reconcile final changes
Pause source writes and validate coverage and results.
Change the writer
Allow the intended writer and validate the application.

The arrows show decisions in order, not guaranteed durations or automatic success. First define what moves and which requirements must remain satisfied.

Attributes to check even when both sides are SQL Server
CheckA common gap
Engine version, edition and DB compatibilityChanging compatibility level does not make a newer-engine backup restorable on an older engine
Database data, schema and featuresVerify types, collation, encryption and used features for the method and destination
Server-level dependenciesInventory logins, SQL Agent jobs, linked systems and file/OS dependencies separately
ApplicationsInclude drivers, TLS, identity, destinations, timeouts and business behavior

Choose initial copying and ongoing replication separately

Methods to evaluate for ordinary RDS for SQL Server
CandidateRequirements and checksStill separate
Native backup/restoreDatabase-level copy; check S3, options, versions, backup sequence and time needed while writes are pausedServer logins/jobs, applications, final data and traffic cutover
AWS DMS full load plus CDCA candidate for tracking changes after initial loading; verify versions, access, logs and type/feature restrictionsObject preparation, replication coverage, final changes and business results
Native initial copy plus DMSA hybrid candidate; plan the handoff position and data continuityCombining tools does not automatically establish consistency

Change data capture (CDC) tracks ongoing changes so they can be applied to a destination. Connectivity alone does not establish eligibility. Verify source and target prerequisites and the scope of tables, changes and schema. DMS is not a copy of an entire server environment.

Example M1: compare a 15-minute write-pause budget

The fictional CorpDocs moves from self-managed SQL Server 2022 Standard to an eligible RDS for SQL Server 2022 Standard destination. The durations are supplied teaching values, not measured AWS performance. For this arithmetic comparison, assume method eligibility and feature dependencies have been resolved. A real project must first verify those assumptions.

Add the stages required after writes are paused
MethodPaused stages, minutesTotalAgainst the supplied 15-minute limit
Offline full backup/restoreBackup 2 + transfer 3 + restore 8 + validation 5 + switch 220 minExceeds the stipulated limit
Precopy with ongoing replication, then cut overDrain final changes 3 + validate 5 + switch 210 minWithin the given budget, subject to eligibility and acceptance

Example M1: change 105 has not arrived yet

Correlate observations before and after changing the writer
StateSourceDestinationDecision
Initial copy completedChanges through 105 occurred while runningApplied through 104105 is missing; cutover conditions fail
Final acceptanceSource writes paused; final change fixed at 105Applied through 105; required data/access/business checks passA candidate within the stated checks; designate one writer
After cutoverRemains at 105New destination-only change 106The databases now differ; returning the endpoint cannot copy 106 back

104, 105 and 106 are teaching change numbers, not SQL Server log positions or displayed DMS metrics. Real acceptance must account for final transactions, related tables, errors, unapplied work and business results. Equal row counts or an available status alone do not establish equal contents.

Plan cutover and fallback around write ownership

  1. Record scope, eligibility, jobs/logins/external dependencies and acceptance criteria.
  2. Define source-write and in-flight-work handling, then reconcile the final source change with the destination.
  3. Verify data, source network, TLS, identity, access and required read/write/business results.
  4. Deliberately redirect applications and keep one business writer; include connection pools and other jobs.
  5. Define fallback authority, decision deadline, writes to pause and how post-cutover data would be reconciled.

Check the migration decisions

Choose an option to see the result and explanation.

M1 offline stages are 2 + 3 + 8 + 5 + 2 minutes. Do they fit 15 minutes?
Read the answer and reason

Answer: No; the total is 20 minutes, beyond the supplied limit. · Compare the sum of all stages with 15; this does not predict real AWS timing.

The initial copy is complete. Source is at 105; target has applied 104. What next?
Read the answer and reason

Answer: Control source writes and verify 105 and the required acceptance checks. · Reconcile final changes, coverage and data/business acceptance before deciding.

After cutover, only the target accepted 106. Does switching the endpoint back also return that data?
Read the answer and reason

Answer: No; plan how 106 will be retained and reconciled. · Changing a destination does not copy target-only data back.

A source job reads local files using PowerShell. Is a DB restore to ordinary RDS sufficient?
Read the answer and reason

Answer: No; inspect supported steps and replacement processing. · Do not reject all SQL Agent work; isolate unsupported PowerShell and the actual dependency.

Check, then explain in your own words

  1. Identify source, destination and feature dependencies.
  2. Sum all paused stages and compare with the allowed write pause.
  3. Reconcile final changes with destination acceptance.
  4. Explain fallback including destination-only writes.
Check your understanding: DMS full load completed. Is application migration therefore complete?

Not yet. Ongoing changes, final coverage, dependencies, applications and the writer still need acceptance.

One data-transfer stage and the acceptance criteria for moving the business workload have different scopes.

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.