Question 260
During a production change postcheck at fictional Marble Bay Software, the change record states: A gNMI telemetry collector cannot process bursty device messages and drops interface counters without reporting any gaps. Which sustainable preventive control most directly avoids a recurrence?
Explanation
Correct Answer
Measure ingress rate, queue depth, drop rate and end-to-end sample latency with capacity alerts. This prevents the recurrence because streaming telemetry pipelines have collectors, transports, ingestion queues and storage capacity that can fail independently of device subscriptions. A matching immediate correction is: Introduce bounded buffering, backpressure monitoring and horizontal ingestion capacity with visible gap or loss reporting. Applied here, the relevant facts are: A gNMI telemetry collector cannot process bursty device messages and drops interface counters without reporting any gaps. The question asks which sustainable preventive control most directly avoids a recurrence.
Incorrect Answers
That safeguard is relevant to: cA-signed certificates are public trust assertions but the associated private key must remain confidential for endpoint authentication. The current failure instead follows from: streaming telemetry pipelines have collectors, transports, ingestion queues and storage capacity that can fail independently of device subscriptions. Its prevention measure is “Measure ingress rate, queue depth, drop rate and end-to-end sample latency with capacity alerts”.
This routine guards against another mechanism: webhooks are asynchronous network events; integrity, delivery acknowledgements and idempotent processing are separate reliability needs. The recurring exposure in the current case is: streaming telemetry pipelines have collectors, transports, ingestion queues and storage capacity that can fail independently of device subscriptions. The matching preventive check is “Measure ingress rate, queue depth, drop rate and end-to-end sample latency with capacity alerts”.
The suggested review tracks a different technical risk: pyATS can learn and compare device operational state to detect unintended network effects beyond text config diffs. Here the sustained control must address: streaming telemetry pipelines have collectors, transports, ingestion queues and storage capacity that can fail independently of device subscriptions. Choose “Measure ingress rate, queue depth, drop rate and end-to-end sample latency with capacity alerts”.