Error handling
When a failure happens in a Power Automate cloud flow, the default behavior is to stop processing. You might want to handle errors and roll back earlier steps in case of failure.
A solution architect should specify how errors are to be handled within cloud flows.
Run after
The way that errors are handled is by changing the run after settings in the steps in the flow, as shown in the following image.
Parallel branches
When using the run after settings, you can have different actions for success and failure by using parallel branches.
Changesets
If your flow needs to perform a series of actions on Dataverse data, and you must ensure that all steps work or none of them work, then you should use a changeset.
If you define a changeset, the operations will run in a single transaction. If any of the steps error, the changes that were made by the prior steps will be rolled back.
Note
Changesets are only available with the Create, Update, and Delete actions in the Dataverse connector.
API limits
The platform has API limits and service protection limits. Power Automate observes these limits; each step in a flow consumes an API call.
Limits are also applied for certain actions in Power Automate and for Power Automate:
- Many operations like the Apply for Each loop only work up to 100,000 iterations. You might be required to partition work when working with large sets of items.
- The Do Until loop has a default of 60 loops or 1 hour of run time. If it exceeds these limits, it exits the loop without error. You can increase the number of loops and time in the settings for the action.
- Flows can run for a maximum of 30 days. You should not use long running flows; use scheduled flows that check if the row needs to be processed instead.
- Connectors have throttling limits, for instance the Dataverse connector permits only 6,000 API calls for each connection every 300 seconds.