Salesforce Certified MuleSoft Integration Architect (MCIA) - Practice Questions & Exam Prep

A Mule application must guarantee that an order message published to Anypoint MQ is processed exactly once, even if a consumer worker crashes mid-processing and the message is redelivered. Which design most reliably enforces exactly-once processing semantics?

• A. Increase the Anypoint MQ message lock TTL so the message is never redelivered.
• B. Persist a processed-message id in a shared Object Store and check it with an idempotent receiver before doing the work.
• C. Set the consuming flow's processing strategy to synchronous so only one thread runs.
• D. Wrap the consumer logic in an Until Successful scope with a high retry count.

Correct: B

Anypoint MQ guarantees at-least-once delivery, so redelivery is always possible; the architect must add idempotency at the application layer. A shared (not in-memory) Object Store keyed by a stable message id lets every worker detect a duplicate before repeating side effects. A longer lock TTL only delays redelivery, a synchronous strategy does not survive a crash, and Until Successful retries the same work rather than deduplicating it.

An integration must aggregate data from three independent backend APIs and return a single response. Two of the three calls can run in parallel; the third depends on the result of the first. The architect wants the lowest end-to-end latency without losing any partial-failure visibility. Which approach fits best?

• A. Call all three APIs sequentially inside a single flow so error handling stays simple.
• B. Use a Scatter-Gather to call all three APIs concurrently and merge the results.
• C. Run the first call, then a Scatter-Gather for the second and dependent third call.
• D. Call the first API, then run the second and third concurrently (e.g. Scatter-Gather), and inspect each route's result for partial failures.

Correct: D

The dependency forces the first call to complete before the dependent call can start, so it cannot be part of the same concurrent group. Running the independent second call and the dependent third call concurrently after the first minimizes latency, and Scatter-Gather aggregates per-route results (including failures) so partial-failure visibility is preserved. Pure sequential execution (A) is slowest; gathering all three at once (B) violates the dependency; option C pairs the wrong two calls in parallel.

A customer requires that a Mule integration survive the loss of an entire CloudHub availability zone with no manual intervention and minimal data loss. Which architectural choice most directly meets this requirement?

• A. Deploy a single worker with the largest available vCore size.
• B. Deploy multiple workers across availability zones and use persistent, externally durable state rather than in-memory state.
• C. Schedule a nightly backup of the worker's local disk.
• D. Enable verbose logging so failures can be replayed manually after an outage.

Correct: B

Surviving an AZ loss without manual intervention requires horizontal redundancy across zones plus state that does not live only inside a worker; CloudHub spreads multiple workers across AZs, and durable external state (e.g. persistent Object Store or a database) prevents data loss when a worker disappears. A single large worker (A) is a single point of failure, disk backups (C) and verbose logging (D) are recovery aids, not automatic resilience.