Cardinality problem on source of type 'fluent' #13434
Labels
domain: observability
Anything related to monitoring/observing Vector
source: fluent
Anything `fluent` source related
type: bug
A code related bug.
We are running vector 0.22.2 with several sources enabled (fluent, http, syslog, tcp) and are experiencing metrics cardinality problems on our fluent source. We have instances delivering logs via fluent protocol to vector via fluentbit/td-agent-bit. When retrying connections after an idle_timeout or when restarting an instance metrics are produced for the source_host:port pair. That way we up to thousands of metrics for a single host as the source port changed. That leads to a tremendous number and volume of metrics scraped by prometheus.
We would like to see metrics for fluentd configurable in that way, that including the source port is either an optional feature or that we can suppress getting metrics on a source_host:source_port basis by a configuration option giving us metrics only for the source_host.
The text was updated successfully, but these errors were encountered: