etl
2 TopicsWhat Are the Ways to Dynamically Invoke Pipelines in ADF from Another Pipeline?
I am exploring different approaches to dynamically invoke ADF pipelines from within another pipeline as part of a modular and scalable orchestration strategy. My use case involves having multiple reusable pipelines that can be called conditionally or in sequence, based on configuration stored externally (such as in a SQL Managed Instance or another Azure-native source). I am aware of a few patterns like using the Execute Pipeline activity within a ForEach loop, but I would like to understand the full range of available and supported options for dynamically invoking pipelines from within ADF. Could you please clarify the possible approaches for achieving this? Specifically, I am interested in: Using ForEach with Execute Pipeline activity How to structure the control flow for calling multiple pipelines in sequence or parallel. How to pass pipeline names dynamically. Dynamic pipeline name resolution Is it possible to pass the pipeline name as a parameter to the Execute Pipeline activity? How to handle validation when the pipeline name is dynamic? Parameterized execution Best practices for passing dynamic parameters to each pipeline when calling them in a loop or based on external config. Calling ADF pipelines via REST API or Web Activity When would this be preferred over native Execute Pipeline? How to handle authentication and response handling? If there are any recommendations, gotchas, or best practices related to dynamic pipeline orchestration in ADF, I would greatly appreciate your insights. Thanks!25Views0likes0CommentsHow to Orchestrate ADF Pipelines as Selectable Steps in a Configurable Job
I am working on building a dynamic job orchestration mechanism using Azure Data Factory (ADF). I have multiple pipelines in ADF, and each pipeline represents a distinct step in a larger job. I would like to implement a solution where I can dynamically select or deselect individual pipeline steps (i.e., ADF pipelines) as part of a job. The idea is to configure a job by checking/unchecking steps, and then execute only the selected ones in sequence or based on dependencies. Available resources for this solution: Azure Data Factory (ADF) Azure SQL Managed Instance (SQL MI) Any other relevant Azure-native service (if needed) Could you please suggest a solution that meets the following requirements: Dynamically configure which pipelines (steps) to include in a job. Add or remove steps without changing hardcoded logic in ADF. Ensure scalability and maintainability of the orchestration logic. Keep the solution within the scope of ADF, SQL MI, and potentially other Azure-native services (no external apps or third-party orchestrators). Any design pattern, architecture recommendations, or examples would be greatly appreciated. Thanks!22Views0likes0Comments