sqlpackage is a Microsoft provided command-line utility that automates a number of database development tasks by exposing some of the public .NET based Data-Tier Application Framework (DacFx) APIs.
Microsoft promotes using
sqlpackage with a .bacpac file in their documentation for moving databases to Azure SQL Database, but mostly just share a couple of command line samples, and leave the gory details to the reader. Hopefully this post can help you be successful with this approach to moving databases in (and out) of Azure SQL Database.
Setting the stage
A customer wanted to move a SQL Server 2008 database hosted on a Windows Server 2008 VM to Azure SQL Database (single database) for archival purposes.
The database was not in use, so no requirements for availability during the move - so "offline" migration.
The customer wanted to move all tables.
Fairly large database, about 700 GB, with largest table consuming 500 GB with 4 billion rows.
Microsoft lists a number of options/tools, and each one was considered:
Azure Database Migration Service / Transactional replication was ruled out - no need for online migration and the extra work involved (and it requires schema changes).
Azure Data Factory / Bulk Copy operates on single tables, too many moving parts.
Azure Data Sync - online sync not needed.
Data Migration Assistant - concerned about size and volume.
BACPAC with SqlPackage seemed like a good fit based on the official statements below. You can read more about the .bacpac file format here
For scale and performance with large databases sizes or a large number of databases, consider using the SqlPackage command-line tool to export and import databases
We encourage using SqlPackage to import/export databases larger than 150 GB
We took the approach depicted here, based on a restored backup of the original database.
First the database and table sizes were investigated, using a script like this to collect disk usage data:
SELECT t.NAME AS TableName, i.name as indexName, p.[Rows], sum(a.total_pages) as TotalPages, sum(a.used_pages) as UsedPages, sum(a.data_pages) as DataPages, (sum(a.total_pages) * 8) / 1024 as TotalSpaceMB, (sum(a.used_pages) * 8) / 1024 as UsedSpaceMB, (sum(a.data_pages) * 8) / 1024 as DataSpaceMB FROM sys.tables t INNER JOIN sys.indexes i ON t.OBJECT_ID = i.object_id INNER JOIN sys.partitions p ON i.object_id = p.OBJECT_ID AND i.index_id = p.index_id INNER JOIN sys.allocation_units a ON p.partition_id = a.container_id WHERE t.NAME NOT LIKE 'dt%' AND i.OBJECT_ID > 255 AND i.index_id <= 1 GROUP BY t.NAME, i.object_id, i.index_id, i.name, p.[Rows] ORDER BY object_name(i.object_id)
Then the database was analyzed using the Data Migration Assistant.
Based on the analysis results, it turned out that multiple functions, stored procedure and views were referencing other databases. It was decided to drop all these objects, as only the table data would be needed going forward. Alternative would have been to refactor these objects.
The following script was used to drop all stored procedures, and similar scripts were run to remove views and functions.
DECLARE @sql VARCHAR(MAX) = '' , @crlf VARCHAR(2) = CHAR(13) + CHAR(10); SELECT @sql = @sql + 'DROP PROCEDURE ' + QUOTENAME(SCHEMA_NAME(schema_id)) + '.' + QUOTENAME(p.name) +';' + @crlf FROM sys.procedures p PRINT @sql; EXEC(@sql);
Database export to .dacpac
Our initial attempt was to simply install the latest .NET Framework version of the SqlPackage MSI and run a .bacpac export using the following commands.
SqlPackage.exe is located in the
C:\Program Files\Microsoft SQL Server\160\DAC\bin folder.
>> cd C:\Program Files\Microsoft SQL Server\160\DAC\bin > SET SET TEMP=E:\TEMP > SET TMP=E:\TEMP >sqlpackage /Action:Export /SourceConnectionString:"Data Source=.\SQLEXPRESS;Initial catalog=AdventureWorks2008R2;Integrated Security=true" /TargetFile:E:\temp\MyDB.bacpac /OverwriteFiles:true /p:CommandTimeout=0 /p:LongRunningCommandTimeout=0 /p:Storage=File /p:TempDirectoryForTableData=E:\TEMP
Notice that all temp files and output files are redirected to a disk with plenty room, in this sample the E: drive.
/p:CommandTimeout=0 /p:LongRunningCommandTimeout=0 sets an indefinite timeout for database queries run during the export.
/p:Storage=File Specifies how elements are stored when building the database model. For performance reasons the default is InMemory. For very large databases, File backed storage is required.
We got this error:
Error SQL71627: The element User: [MYPC\Administrator] has property AuthenticationType set to a value that is not supported in Microsoft Azure SQL Database v12.
So we dropped all Windows users from the database and the export ran well. The resulting .bacpac file was "only" 15 GB in size.
Import .bacpac to Azure SQL Database
As our VM is in Azure, we can just use the .bacpac directly from disk. Alternative is to move the file to an Azure VM and run SqlPackage from that (or copy to blob storage and use the portal based Import feature, see below).
>> cd C:\Program Files\Microsoft SQL Server\160\DAC\bin > SET SET TEMP=E:\TEMP > SET TMP=E:\TEMP >sqlpackage /Action:Import /TargetConnectionString:"Data Source=myazureserver.database.windows.net;Initial Catalog=azuredb;Authentication=Active Directory Default;" /SourceFile:E:\temp\MyDB.bacpac /p:CommandTimeout=0 /p:LongRunningCommandTimeout=0 /p:Storage=File
Using the connection string above will allow you to use Multi-Factor Authentication (MFA) and other supported Azure AD authentication options.
For a long running import process, Microsoft support has created a script to monitor progress
Upload to Azure Storage account
For smaller databases, you can use the Import database feature in the Azure Portal. In order to do this, the .bacpac file must be uploaded to blob storage.
You can use the
azcopy command line tool to upload the .bacpac file to blob storage.
Navigate to your Azure Storage account in the Portal, and generate a short lived access token, then grab the access token URL. Ensure that the token URL contains the container name.
Then use the token URL with AzCopy:
>azcopy copy "MyDB.bacpac" "https://mystorageaccount.blob.core.windows.net/mycontainer?sp=racw&st=2022-06-20T10:35:02...."
You can then use the portal or Azure command line tools to do the import. Notice that this method does not support MFA.
The command line to import from blob storage (PowerShell or Azure CLI) is documented here.