After a Function outage, Event Grid retries create a retry storm into the city’s AI handler. What mitigation helps?
Select an answer to reveal the explanation.
Short Explanation
Retry storms are a jammed doorbell—backoff, fix the Function, and let dead-letter catch what still fails instead of ringing forever.
Full Explanation
After a Function outage, Event Grid retries can stampede the city's AI handler. Mitigation combines exponential backoff and healthy endpoint recovery, sufficient Function capacity, and dead-letter destinations for events that still fail after retries—taming the storm without discarding civic events wholesale when ordinance uploads and case signals resume.
Disabling all Event Grid retries forever and dropping failed events silently loses ordinance uploads and case signals during the outage window. Removing dead-letter destinations leaves exhausted deliveries with nowhere to go for later replay. Scaling the broken Function to zero removes capacity so retries fail harder and prolong the municipal backlog.
Exam caveat: retry storms are a known event-driven failure mode; platforms retry for resilience, but handlers must be idempotent and ops must restore health quickly. Dead-lettering complements retries—it does not replace fixing the endpoint or ensuring scale for catch-up traffic.
Operational check: simulate handler downtime, observe retry deliveries, restore a healthy Premium/scaled endpoint with idempotent processing, confirm backoff behavior in delivery attempts, and verify leftover failures land in dead-letter for controlled replay rather than infinite hammering.