Applies toSQL Server 2016+ on Amazon EC2 to Amazon RDS for SQL Server (Standard / Enterprise / Web)
1. Why move
| On EC2 you do | On RDS |
|---|---|
| OS + SQL patching | AWS-managed minor versions in maintenance windows |
| AG/FCI design and testing | Multi-AZ with automatic failover (synchronous) |
| Backup jobs, retention, PITR | Automated backups and PITR (up to 35 days), snapshots |
| Monitoring agents | CloudWatch, Database Insights, Enhanced Monitoring |
| Storage resizing | Storage autoscaling |
2. Readiness: RDS limits to check
- No
sysadmin, no OS access; some server-level settings via parameter groups only. - Max databases per instance (depends on instance class/edition; check current quota).
- Features delivered as options: SSIS, SSRS, SSAS, native backup/restore, TDE, MSDTC.
- Linked servers: supported to other SQL Servers and some sources (check docs); OLEDB providers you install yourself aren't possible.
xp_cmdshell, CLR (depends on version), FILESTREAM: not supported → keep on EC2 or RDS Custom.
Scan instance features on the EC2 server with the assessment queries (linked servers, CLR assemblies, agent job steps using CmdExec/PowerShell).
3. Size from real data
Use 2–4 weeks of CloudWatch EC2 metrics and SQL counters (CPU p95, memory, EBS IOPS/throughput). RDS Multi-AZ doubles compute cost, but replaces your second AG node — compare like for like.
4. Migrate
| Method | Downtime |
|---|---|
Native backup to S3 → rds_restore_database with NORECOVERY + logs → rds_finish_restore | Final log + finish (minutes) |
| AWS DMS full load + CDC | Minutes (good when consolidating many databases into fewer RDS instances) |
Since both sides are in AWS, use S3 in the same region and a gateway VPC endpoint — fast and no internet traffic.
5. Replace instance-level pieces
- Agent jobs with
CmdExec/PowerShell steps → Lambda/EventBridge Scheduler or SSM Run Command. - Server-level triggers, custom error log handling → CloudWatch Logs + alarms.
- Windows logins → join RDS to AWS Managed Microsoft AD.
- Maintenance (index/statistics) → keep Ola Hallengren scripts as SQL Agent jobs on RDS (supported).
6. Cutover and decommission
- Cutover with the runbook; switch the DNS alias (Route 53 private hosted zone CNAME) to the RDS endpoint.
- Keep EC2 stopped (not terminated) for a rollback period; snapshot volumes.
- Terminate EC2, delete volumes and old snapshots; adjust Savings Plans / RIs.
Tip
Use a Route 53 CNAME like
sales-db.internal.corpfor the app connection. Moving from EC2 to RDS — and later to a bigger RDS or Aurora — becomes a DNS update.