What we ended up doing is:
On the build side we generate an idempotent database creation script:
dotnet ef migrations script --idempotent --output migrations.sql --context ApplicationContext
Where ApplicationContext is the name of your EF context and migrations.sql is the sql script file name.
Then on deployment side we have a small powershell script that effectively runs the migrations.sql script
param(
[Parameter(Mandatory)]
[string]$server,
[Parameter(Mandatory)]
[string]$dbname,
[Parameter(Mandatory)]
[string]$dbadmin,
[Parameter(Mandatory)]
[string]$dbpassword,
[Parameter(Mandatory)]
[string]$migrationPath
)
function Deploy-Migrations ($migrationPath,$DBSettings)
{
#Setting up database connection
$connection = New-Object System.Data.SqlClient.SqlConnection
$connection.ConnectionString = [string]::Format("Data Source=tcp:{0}.database.windows.net,1433;Initial Catalog={1};User Id={2}@{0};Password={3};MultipleActiveResultSets=True", $DBsettings['sqlServerName'], $DBsettings['databasename'],$DBsettings['adminAccount'], $DBsettings['adminPassword'])
try
{
$connection.Open();
$SqlCmd = New-Object System.Data.SqlClient.SqlCommand
$SqlCmd.Connection = $connection
$query = Get-Content $migrationPath
$sqlCmd.CommandText = $query.Replace("GO","") # This is required to prevent "syntax" complaints
$sqlCmd.ExecuteNonQuery()
Write-Host "Migration Deployed"
}
Catch
{
Write-Error "oops ... PAnic ... $($_.Exception.Message) on $($_.Exception.ItemName)"
break
}
Finally
{
$connection.Close()
}
}
$DBSettings = @{"sqlServerName"=$server; "databasename"=$dbname; "adminAccount"=$dbadmin; "adminPassword"=$dbpassword }
Deploy-Migrations $migrationPath $DBSettings