Question 860
A team wants an analytics alert if interface utilization deviates from baseline while still retaining raw evidence. Which shortcut would be most likely to undermine this objective?
Explanation
Correct Answer
This is the unsafe shortcut to avoid: trigger incidents from a single unvalidated stale sample while discarding the source telemetry. It violates the principle that IBA stages can enrich and aggregate telemetry before evaluating conditions that constitute an anomaly. An appropriate alternative is to configure the supported aggregation and alert conditions with documented baseline behavior.
Incorrect Answers
The option “Disable interface telemetry to suppress alerts caused by rising packet errors” is useful in a different setting: A monitoring team wants to detect rising discards on the uplinks of one specific leaf group. The case being tested is: A team wants an analytics alert if interface utilization deviates from baseline while still retaining raw evidence. IBA stages can enrich and aggregate telemetry before evaluating conditions that constitute an anomaly. The directly relevant unsafe shortcut is “Trigger incidents from a single unvalidated stale sample while discarding the source telemetry”.
“Raise the anomaly threshold to infinity because the probe currently selects unrelated devices” is relevant to a different condition: A built-in analytics probe reports a metric but cannot identify why the value changed after a rack migration. The stated question instead concerns: A team wants an analytics alert if interface utilization deviates from baseline while still retaining raw evidence. IBA stages can enrich and aggregate telemetry before evaluating conditions that constitute an anomaly. The directly relevant unsafe shortcut is “Trigger incidents from a single unvalidated stale sample while discarding the source telemetry”.
That choice—“Disable all analytics probes because several symptoms share the same source device”—addresses this other problem: A site sees hundreds of anomaly events from one failing uplink and wants to distinguish one root cause from many symptoms. Here the scenario is: A team wants an analytics alert if interface utilization deviates from baseline while still retaining raw evidence. IBA stages can enrich and aggregate telemetry before evaluating conditions that constitute an anomaly. The directly relevant unsafe shortcut is “Trigger incidents from a single unvalidated stale sample while discarding the source telemetry”.