none
ADFv2 - Deploying to EXISTING production data factory via ARM Template

    Question

  • I have two data factories, let's call them test-dfact and prod-dfact. I have finalised my changes (e.g. renamed pipeline) in test-dfact and exported the ARM Template. I edited the parameters accordingly in arm_template_parameters.json for the production environment, and I have then initiated "Import ARM Template" from prod-dfact. The deployment succeeds, but prod-dfact now has all the resources it did before, plus anything new in the ARM Template. In particular, I renamed a pipeline in test-dfact but it just got added as a new pipeline in prod-dfact.

    What are the recommended practices for this type of deployment to a production environment? Am I supposed to delete the entire production data factory prior to deploying the ARM Template with changes from dev/test? That doesn't seem wise to do in production!

    Wednesday, August 29, 2018 6:06 AM

All replies

  • Hello,

    I believe you are seeing the behavior of ARM incremental mode, which is the default mode for ARM.  Take a look at this article which explains the different modes, as well as other best practices you might find useful:

    https://blogs.msdn.microsoft.com/mvpawardprogram/2018/05/01/azure-resource-manager/ 

    Wednesday, August 29, 2018 5:08 PM
    Moderator
  • Hi Jason,

    Thanks for your response. Won't deploying in "Complete" mode result in losing monitoring details of past pipelines? That was indicated by someone here: https://github.com/MicrosoftDocs/azure-docs/issues/14190#issuecomment-417104051

    Their recommendation was to use a script to rename or delete resources as necessary. I haven't however had a chance to test this approach so I'm still unsure of what is the best workflow to use in continuous improvement with ADF.

    Monday, September 3, 2018 9:07 AM