Network & Infrastructure / Telemetry, Capacity & Incident Analysis
Per-queue congestion hidden by low port utilization
A port averages 30 percent utilization while a critical class drops packets. Its queue may have a small allocation or face bursts independent of the port-wide average.
Environment and inputs: Network and infrastructure observability; Prometheus examples describe metric semantics rather than a universal exporter schema. Diagnostic/design example; not tested on production equipment.
Collect the evidence
Record queue-specific occupancy, drops, scheduler allocation and traffic classification over relevant intervals. Compare counters at the actual congestion point rather than only at the access interface.
Explain the result
Port capacity is shared through scheduling and buffering policy. One class can exceed its available service or burst budget even when the aggregate interface has spare capacity.
Plan the fix
Correct classification or scheduler allocation according to measured service requirements, preserving other classes and control traffic. Do not assume increasing link speed is the only meaningful remedy.
Check the result and rollback
Validate mixed-class workloads and short bursts, including the protected application's latency. Restore the prior scheduler if another required class loses service or the change only moves congestion downstream.