Replies: 1 comment
-
Thanks @renepajta for the summary! The event-based triggers are the best approach to model dependencies between pipelines that are running within different workspaces. As mentioned, you can use storage events or custom events to trigger pipelines. There is also the option to publish events using Web Activity and MSI Authentication: We have a sample of this running in our environment and we are currently evaluating how we can best integrate this into the landing zones as there are service limitations which are not making it an easy decision. |
Beta Was this translation helpful? Give feedback.
-
Scenario: Data Product consumes data from other data product when it's ready
Data Integrations are responsible only to bring the data to Enriched layer from where others can start consuming. Additionally, data products may expose their products for consumption.
There are different approaches how data product could consume data and it should be defined in data contract:
I especially like the last approach as it allows to scale across the organisation. Data Consumers can listen to the event and consume newly arrived data as soon as the load is done. Also, in some cases, data consumer wants to wait until the whole dataset (multiple entities) are loaded before consuming (e.g. loading to start schema).
Synapse Pipelines / Data Factory offers Storage Account events and Custom events to configure event triggers which allows to go this direction.
However, there is little to no integration to publish events from pipeline directly. Therefore, I would be interested in concepts / your experience how we can publish events (I am exploring Web activity or using Azure Functions).
Beta Was this translation helpful? Give feedback.
All reactions