This article is contributed. See the original author and article here.
In the Summer of 2018, we released SQL Server Integration Services (SSIS) in Azure Data Factory (ADF), which is a Platform as a Service (PaaS) for lifting & shifting on-premises SSIS workloads to the cloud. It offers SSIS Integration Runtime (IR) as a managed and dedicated compute resource to run SSIS packages in ADF. We realized early on that there’s a long road ahead for us to support full backward compatibility.
Our customers’ enthusiasm that greeted the release was curbed by their fear of losing accumulated investments in legacy SSIS packages and entrenched operational tools that they use to design, debug, deploy, configure, and manage those packages, as well as to trigger, schedule, or orchestrate their executions. Redesigning hundreds to thousands of those existing packages and retraining their teams to operate new tools are simply non-starters.
Recently, we’ve reached the end of that proverbial long road and can finally allay their fear of losing investments by providing many enablers that unblock and accelerate their legacy SSIS migrations into ADF. On top of that, we’ve also amplified their fear of missing out by offering cloud-first/cloud-only features that allow them to modernize their SSIS workflows in ADF.
This article highlights some of those enablers/features that avoid or at least minimize any changes to their existing packages and tools.
Design, Management, and Migration Tools
One of the main requirements from our customers is to preserve their skillsets in using the familiar tools, such as SQL Server Data Tools (SSDT) and SQL Server Management Studio (SSMS), to design, debug, deploy, configure, execute, and manage their SSIS packages.
To fulfil this requirement, we’ve Azure-enabled both SSDT and SSMS for you to operate on your packages in the cloud just like you’ve done on premises. On top of that, we’ve turned Azure-enabled SSDT into an all-in-one SSIS migration tool where you can perform the 3-step iterative process of package assessment for cloud compatibility, package execution testing in Azure, and package redeployment to Azure.
For more info, see the following docs:
https://docs.microsoft.com/azure/data-factory/how-to-invoke-ssis-package-ssdt
Deployment Models
For a bit of history, SSIS on premises has supported two deployment models for a while:
- Package Deployment Model that has been around since the early days of SSIS. It allows you to deploy packages into file system or SQL Server database (MSDB), and then manage them using the legacy SSIS package store interface on SSMS.
- Project Deployment Model that was introduced with SQL Server 2012. It allows you to deploy projects/packages into SSIS catalog (SSISDB) hosted by SQL Server, and then manage them using the Integration Services Catalogs interface on SSMS.
In the beginning, SSIS in ADF supported only Project Deployment Model where SSISDB is hosted by Azure SQL Database (DB)/Managed Instance (MI). Over time, we’ve learned that converting SSIS packages from Package Deployment Model into Project Deployment Model can be quite challenging, especially when you have to redesign many of them to replace configuration files with project/package parameters. On top of that, our on-premises telemetry shows that SSIS instances with Package Deployment Model continue to outnumber those with Project Deployment Model by two to one.
Now, SSIS in ADF also supports Package Deployment Model where packages can be stored in file system, Azure Files, or MSDB hosted by Azure SQL MI, and managed using the new SSIS IR package store interface on SSMS. This also turns SSISDB from a potential migration blocker into an optional enhancement and makes the provisioning of SSIR IR much simpler, because bringing your own Azure SQL DB/MI to host SSISDB is not mandatory anymore.
With Project Deployment Model, you can use Integration Services Deployment Wizard on SSDT/SSMS to deploy projects/packages into SSISDB hosted by Azure SQL DB/MI. With Package Deployment Model, you can use dtutil command prompt utility to deploy packages into Azure Files or MSDB hosted by Azure SQL MI, while switching their protection level at the same time.
For more info, see the following docs:
https://docs.microsoft.com/azure/data-factory/tutorial-deploy-ssis-packages-azure#creating-ssisdb
https://docs.microsoft.com/azure/data-factory/azure-ssis-integration-runtime-package-store
Orchestration Tools
On premises, most of our customers use SQL Server Agent to orchestrate SSIS package executions on SQL Server. Their packages can be stored in file system or MSDB/SSISDB hosted by SQL Server and referenced directly or indirectly via SSIS package stores.
If you migrate your SQL Server to Azure SQL MI, you can use Azure SQL MI Agent, which is a built-in tool similar to SQL Server Agent, to orchestrate SSIS package executions on SSIS IR in ADF. Your packages can be stored in file system, Azure Files, or MSDB/SSISDB hosted by Azure SQL MI and referenced directly or indirectly via SSIS IR package stores.
Since Azure SQL MI Agent supports SSIS jobs via SSISDB hosted by Azure SQL MI, if you want to use it as your orchestration tool, we’ll need to create and manage SSISDB on your behalf, even if you use Package Deployment Model that doesn’t store packages in SSISDB.
For more info, see the following doc:
https://docs.microsoft.com/azure/data-factory/how-to-invoke-ssis-package-managed-instance-agent
If you migrate your SQL Server to Azure SQL DB, there’s no built-in tool similar to SQL Server Agent to orchestrate SSIS package executions on SSIS IR in ADF. To address this gap, we’ve provided SSIS scheduling feature on SSMS for you to auto-create new ADF pipelines, activities, and triggers that orchestrate SSIS package executions on SSIS IR. We’ve also provided SSIS Job Migration Wizard on SSMS for you to convert your existing SSIS jobs created for SQL Server Agent into new ADF pipelines, activities, and triggers that orchestrate SSIS package executions on SSIS IR.
For more info, see the following docs:
https://docs.microsoft.com/azure/data-factory/how-to-migrate-ssis-job-ssms
If you use a third-party tool, such as Tivoli, Control-M, ActiveBatch, etc., to orchestrate SSIS package executions on SQL Server, you probably want to continue using it to orchestrate SSIS package executions on SSIS IR in ADF after migrating your SQL Server to Azure SQL DB/MI. To facilitate this, we’ve provided Azure-enabled DTExec (AzureDTExec) command prompt utility that can be configured on SSMS and invoked by your third-party tool. It auto-creates new ADF pipelines and activities for SSIS package executions on SSIS IR that can be orchestrated by your third-party tool.
For more info, see the following doc:
https://docs.microsoft.com/azure/data-factory/how-to-invoke-ssis-package-azure-enabled-dtexec
On-Premises Data Access
After migrating your SQL Server to Azure SQL DB/MI and SSIS workloads to SSIS in ADF, you probably still have data stores on premises that you want to access from packages running on SSIS IR. We’ve provided two access methods: Virtual Network (VNet) injection of SSIS IR and Self-Hosted IR (SHIR) as a proxy for SSIS IR.
With VNet injection, the on-premises network where your data stores reside must be connected via ExpressRoute/VPN Gateway to the VNet where your SSIS IR is injected. In this way, those data stores will get their own private IP addresses that are accessible by your SSIS IR inside the same VNet. This requires you to modify the relevant connection strings to use these private IP addresses. Without redesigning your packages, you can make the above modifications at run-time by assigning the appropriate values to your SSIS parameters, connection manager properties, or property overrides.
With SHIR as a proxy, you can install SHIR in the on-premises network where your data stores reside. In this way, your packages can run on SHIR and still access those data stores just like they did when they ran on SQL Server. This requires you to enable the ConnectByProxy property of relevant connection managers. Without redesigning your packages, you can make the above modifications at run-time by assigning the appropriate values to your SSIS parameters, connection manager properties, or property overrides.
For more info, see the following blog:
Authentication Methods
When running SSIS packages on SQL Server, you probably used Windows authentication to access on-premises data stores like most of our customers. Now that you run SSIS packages on SSIS IR in ADF, you can still access the same data stores using Windows authentication feature. Without redesigning/modifying your packages, you can enable this feature in several ways, for example via Execute SSIS Package activities in ADF pipelines, SSISDB stored procedure to set execution credentials, or Windows cmdkey command as express custom setups.
For more info, see the following doc and blog:
https://docs.microsoft.com/azure/data-factory/ssis-azure-connect-with-windows-auth
If your on-premises data stores were databases on SQL Servers, you probably used Windows authentication via OLEDB connectors to access them like most of our customers. If you’ve migrated those databases to Azure SQL DB/MI, you can’t use Windows authentication to access them anymore, since Azure SQL DB/MI doesn’t support it, but you can use SQL authentication instead. This requires you to modify the relevant connection strings to switch from Windows authentication to SQL authentication. Without redesigning your packages, you can make the above modifications at run-time by assigning the appropriate values to your SSIS parameters, connection manager properties, or property overrides.
Now that you run SSIS packages on SSIS IR in ADF, you can also securely store your SQL authentication credentials in Azure Key Vault (AKV) when you use Execute SSIS Package activities in ADF pipelines.
For more info, see the following doc:
If you want to omit authentication credentials altogether from your packages, you can use Azure Active Directory (AAD) authentication with ADF managed identity to access Azure SQL DB/MI. This requires you to enable the ConnectUsingManagedIdentity property of relevant connection managers and modify the relevant connection strings to use the latest OLEDB driver for SQL Server (MSOLEDBSQL). Without redesigning your packages, you can make the above modifications at run-time by assigning the appropriate values to your SSIS parameters, connection manager properties, or property overrides.
For more info, see the following doc:
Monitoring and Alerting Capabilities
After provisioning your SSIS IR in ADF, you can check its operational status on ADF monitoring hub, where you’ll also find various informational/functional tiles depending on how you’ve configured its general, deployment, and advanced settings.
For more info, see the following doc:
With Project Deployment Model, SSIS package execution logs are always stored in SSISDB internal tables and views, so you can interactively query/analyze and visually present them using designated tools like SSMS. With Package Deployment Model, SSIS package execution logs can be stored in file system or Azure Files as CSV files, but you still need to parse and process them using other designated tools first, before you can interactively query/analyze and visually present them.
Now, SSIS in ADF is also integrated with Azure Monitor, so you can interactively query/analyze and visually present all metrics and logs generated from SSIS IR operations and SSIS package executions on Azure portal, regardless of your deployment models. On top of that, you can also raise near-real time alerts on those metrics and logs.
For more info, see the following doc:
I hope you’ll find the above enablers/features useful to unblock and accelerate your legacy SSIS migration into ADF. Please don’t hesitate to contact us if you have any feedbacks, questions, or issues, and we’ll follow up ASAP. Thank you as always for your support.
Brought to you by Dr. Ware, Microsoft Office 365 Silver Partner, Charleston SC.
Recent Comments