Retry policy
The configurable behaviour of an integration on transient failure — how many retries, with what backoff, and on which error conditions.
A retry policy is the configurable behaviour of an integration step when it encounters a transient failure — network timeout, brief service outage, HTTP 429 throttling, brief connector unavailability. Standard policies include exponential backoff (retry with increasing delays — 1s, 2s, 4s, 8s — to give the dependent service time to recover), fixed-interval retry (retry at constant intervals), and immediate retry (retry without delay; suitable only for very transient issues). Power Automate actions, Logic Apps actions, and Dataverse plug-in registrations all support configurable retry policies. Apply retry only to idempotent operations — non-idempotent ones produce duplicates on retry. After retries exhausted, route to dead-letter handling or alert operations.