Plan for Failures

Tasks can fail because of service outages, expired authorization, missing input, or a changed external resource. Decide how failures are detected and recovered before production use.

During task design

Stop or request review when required data is missing.
Confirm the destination and item count before an external update.
Retry only temporary failures and limit the retry count.
Make repeated runs safe from duplicate records or messages.
Notify an owner when an important run fails.
Use approval or the inbox when a person must decide what happens next.

After a failure

Open Run History and find the first failed step.
Check the connection, target resource, input, and service limits.
Fix the underlying cause.
Use Retry or Resume from failure when available.
Confirm that the external service did not receive a duplicate update.

When escalating, share the task name, run time, failed step, and visible error. Never include credentials or other secrets.