A parking-enforcement feed uses multiple OCI Streaming partitions for throughput. Stakeholders ask whether the entire stream preserves a single global order of every citation. What is the correct order guarantee?
Select an answer to reveal the explanation.
Short Explanation
Order inside one partition is like one teller’s ticket queue—first in stays first. Across many teller windows, there is no single city-wide queue number. Streaming guarantees order within a partition as produced; it does not promise global order across partitions. Total order for the whole stream means one partition and the throughput that implies.
Full Explanation
OCI Streaming guarantees order within a partition in the order messages were produced. It does not guarantee a single global order across all partitions. A parking-enforcement feed that uses multiple partitions for throughput therefore cannot promise one city-wide sequence of every citation. If stakeholders truly need total order for the entire stream, the design must use a single partition and accept the throughput implications. Kafka compatibility does not magically collapse partitions into one ordered log for every consumer.
Claims that Streaming always preserves global order across partitions with exactly-once semantics overstate both ordering and delivery guarantees. Saying order is undefined inside a partition but guaranteed across the whole stream reverses the real rule. Assuming Kafka compatibility automatically collapses all partitions into one ordered log invents a behavior Streaming does not provide.
Exam caveat: per-key order is achieved by keying related messages to the same partition; that is still not global order across keys on different partitions.
Operational check: publish numbered messages with different keys across a multi-partition stream, consume from all partitions, and demonstrate that only same-partition (same-key) sequences stay ordered while cross-partition merge order is not a service guarantee.