Network & Infrastructure / Telemetry, Capacity & Incident Analysis
Unbounded metric labels from per-flow infrastructure telemetry
A monitoring system slows after adding source IP, destination IP and session ID labels to every flow metric. The number of time series grows with traffic diversity rather than stable infrastructure size.
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
Measure active series, label-value growth, ingestion rate, query cost and retention footprint. Identify which labels create combinations that are short-lived or effectively unbounded.
Explain the result
Metric systems typically retain separate series for label combinations. High-cardinality flow identity can consume substantial resources even when the metric name count remains small.
Plan the fix
Keep metrics for bounded aggregate dimensions and route detailed flow evidence to a suitable event or flow system. Preserve the dimensions needed for operational decisions without exporting every identifier as a label.
Check the result and rollback
Load-test expected growth and common queries before broad rollout. Revert the label expansion if ingestion, retention or alert evaluation exceeds capacity or required service monitoring becomes unreliable.