Clarify/adjust MET-001 for expected high-cardinality “inventory” metrics (e.g., kube_pod_info)

[CharlieTLe]

Problem

Rule MET-001 currently states:

Attribute keys on metrics, aggregated by metric name, MUST have less than 10,000 unique values within a 1-hour window.

This appears to flag certain Kubernetes/inventory-style metrics (e.g., kube_pod_info) where a label value is inherently high-cardinality (pod name) and large clusters can exceed 10,000 unique values/hour under normal conditions.

Example

In large clusters, kube_pod_info (or similar “*_info” / inventory metrics) may produce one series per pod, and the pod label can easily exceed the MET-001 threshold within 1 hour, even though this is expected behavior for that metric.

Question / Proposal

Should MET-001 be clarified or adjusted to avoid penalizing expected high-cardinality inventory/entity metrics?

Possible approaches:

1. Scope MET-001 to application/service metrics (exclude infra exporter / inventory metrics).
2. Add explicit exceptions for inventory-style metrics (e.g., *_info) or other clearly-defined categories.
3. Refine the rule to focus on preventing unbounded identifiers on high-volume metrics, while allowing “entity enumeration” metrics.

Desired outcome

• Keep the rule objective and implementable.
• Avoid ambiguous exemptions.
• Add examples of metrics/labels that should/shouldn’t trigger.

[jpkrohling]

@CharlieTLe , would you be interested in opening a PR with the proposed changes?