Proposed title of this feature request

Reduce Vector cardinality

What is the nature and description of the request?

In multitenant OpenShift clusters cluster (different companies accessing to the same cluster or different teams and this could include also to the OSD cluster) where they should not receive in their external tools the logs from the other companies or developers teams, the number of number of inputs, outputs and pipelines in the clusterLogForwarders is big causing a high cardinality in some of the metrics exposed by Vector.

The analysis of the high cardinality is explained in the Red Hat Knowledge Article - https://access.redhat.com/solutions/7137995 section "Diagnostic steps" and this impacts to the OpenShift Monitoring stack, in specific to the memory, cpu and storage required by the Prometheus pods.

It's required to review if it's possible to reduce the high cardinality observed in Vector or removing some metricso or limiting the "tag cardinality" [0]

Why does the customer need this? (List the business requirements)

It impacts to the OpenShift Monitoring stack and the resources required by it: cpu, memory and storage. Any impact to the Monitoring stack, it could impact to the stability of the cluster as the Monitoring stack is required for basic operations in the cluster as: upgrade activities, horizontal scaling, vertical scaling, etc.

List any affected packages or components.

Logging collector Vector

[0]  Vector has a feature, that it's in "status: beta" for now for limiting the "tag cardinality". This feature is documented in https://vector.dev/docs/reference/configuration/transforms/tag_cardinality_limit/ and it's requested to review it for having the option of enabling it and reducing the cardinality.