Migrate SQL Server to Azure SQL Managed Instance with Log Replay Service (LRS)

#When to choose LRS

• No direct network between SQL Server and MI (backups travel via Blob Storage).
• Versions or editions not supported by the MI link.
• You want to use your existing backup jobs (full + diff + log).

Downtime = time to take and restore the last log backup + app switch (usually minutes).

#1. Backup requirements

• Backups with CHECKSUM (required).
• One folder per database in the container (/sales/, /hr/); full, diff and log files for that database only.
• Stripe large databases across multiple files for speed.

BACKUP DATABASE Sales TO URL = 'https://stmig.blob.core.windows.net/backups/sales/sales_full.bak'
  WITH CHECKSUM, COMPRESSION, FORMAT;
BACKUP LOG Sales TO URL = 'https://stmig.blob.core.windows.net/backups/sales/sales_log_0001.trn'
  WITH CHECKSUM, COMPRESSION;

Or back up to disk and upload with azcopy:

azcopy copy "D:\Backups\Sales\*" "https://stmig.blob.core.windows.net/backups/sales/?<sas>" --recursive

#2. Start LRS

Continuous mode (you decide when to complete; best for production cutovers):

az sql midb log-replay start -g rg-sql --mi sqlmi-prod -n Sales \
  --storage-uri "https://stmig.blob.core.windows.net/backups/sales" \
  --storage-sas "<read+list SAS without leading ?>"

Managed identity of the MI can be used instead of a SAS (current versions; grant Storage Blob Data Reader).

Autocomplete mode (all backups already uploaded; finishes automatically):

az sql midb log-replay start ... -a --last-bn "sales_log_0042.trn"

#3. Monitor

az sql midb log-replay show -g rg-sql --mi sqlmi-prod -n Sales

Keep shipping new log backups to the folder; LRS restores them in order.

#4. Cutover

1. Stop the app.
2. Take a final tail-log backup with NORECOVERY on the source (prevents further changes) and upload it.
3. Complete:

az sql midb log-replay complete -g rg-sql --mi sqlmi-prod -n Sales --last-backup-name "sales_log_tail.trn"

4. The database comes online read-write on MI. Switch connection strings; run smoke tests.

#5. Common errors

Warning

While LRS runs, make sure no other log backup job (your backup tool, maintenance plan) takes log backups that don't land in the LRS folder. A missing log file breaks the chain and forces a restart from a new full backup.