Built-in metric registry
Each built-in metric family has a category, a recording path, and a label-granularity scheme. Live scrapes only show series enabled in the active metrics plan — absence from scrape means the family is not in that plan, not that the product lacks the metric.
For scrape/OTLP setup and collect vs emit cost, see Metrics configurability. For per-series increment rules and PromQL, see Built-in metrics.
How to read this table
| Column | Meaning |
| Category | Dataplane category; included via base and categories.include / exclude |
| minimal / standard | Whether the category (hence the family) is on for that base |
| Timing | Hot-path (query workers) vs scrape-time (export refresh) |
| Dimensions | Label keys controlled by granularity presets / overrides (empty = no plan dimensions beyond fixed labels) |
Default collect/emit when a category is on: both true. Override with metrics.collection.<category>.
Opt-in-only families (none on day one) appear here when added; they stay off until explicitly included.
Category membership
| Category | minimal | standard | Typical series |
volume | yes | yes | Queries, responses, truncations, drops, ACL, queries-by-pool |
failures | yes | yes | Parse / forward / script errors, retries, slot exhaustion |
lookup | yes | yes | Lookup outcomes and cache lookups |
timing | no | yes | Duration histograms and forward attempts |
cache_detail | no | yes | Cache fills, singleflight, entries / bytes / evictions / LMDB errors |
forward_detail | no | yes | Forward outstanding gauge (scrape) |
health | yes | yes | Probe results and backend health gauges (included in minimal) |
runtime | no | yes | Slot in-use / capacity gauges |
topology | yes | yes | Configured backends, listener / backend info |
process | no | yes | Process RSS, FDs, threads, CPU (Linux scrape) |
meta | yes | yes | Build info, start time, config generation |
Families by category
volume
| Family |
Timing |
Typical dimensions (fine) |
conduit_queries_total |
hot |
listener, protocol, qtype, qclass, ip_family |
conduit_queries_by_pool_total |
hot |
pool |
conduit_queries_dropped_total |
hot |
listener, protocol, reason, ip_family |
conduit_responses_total |
hot |
listener, protocol, rcode, answer_source, ip_family |
conduit_responses_truncated_total |
hot |
listener, protocol, answer_source, ip_family |
conduit_acl_decisions_total |
hot |
tier, action, listener, ip_family |
Coarse presets drop high-cardinality labels (for example qtype / ip_family). Response rcode values use coarse class buckets or IANA names via granularity.responses.
failures
| Family |
Timing |
Typical dimensions (fine) |
conduit_parse_rejected_total |
hot |
reason |
conduit_forward_errors_total |
hot |
pool, backend, reason |
conduit_retries_total |
hot |
pool |
conduit_script_errors_total |
hot |
reason, script, table |
conduit_slot_pool_exhausted_total |
hot |
|
lookup
| Family |
Timing |
Typical dimensions (fine) |
conduit_lookup_provider_outcomes_total |
hot |
profile, provider, outcome |
conduit_cache_lookups_total |
hot |
cache, profile, result |
timing
| Family |
Timing |
Typical dimensions (fine) |
conduit_phase_duration_seconds |
hot |
phase |
conduit_forward_duration_seconds |
hot |
pool, backend |
conduit_forward_attempts_total |
hot |
pool, backend, outcome |
conduit_lookup_duration_seconds |
hot |
|
conduit_cache_duration_seconds |
hot |
|
conduit_response_duration_seconds |
hot |
|
Pool-only overrides (for example granularity.timing: [pool]) drop backend where applicable.
| Family |
Category |
Timing |
| Cache fill / singleflight / entries / bytes / evictions / LMDB errors |
cache_detail |
hot / scrape as documented in Built-in metrics |
conduit_forward_outstanding |
forward_detail |
scrape |
conduit_probe_results_total, conduit_backend_health_observed, related health gauges |
health |
hot / scrape |
| Slot in-use / capacity gauges |
runtime |
scrape |
conduit_pool_backends_configured, listener / backend info |
topology |
scrape |
conduit_process_resident_bytes, conduit_process_open_fds, conduit_process_max_fds, conduit_process_threads, conduit_process_cpu_seconds_total |
process |
scrape (Linux) |
conduit_build_info, conduit_start_time_seconds, conduit_uptime_seconds, conduit_config_generation |
meta |
scrape |
Exact label sets and PromQL: Built-in metrics.
Event export (separate axis)
| Family prefix |
Controlled by |
Notes |
conduit_events_* |
metrics.event_export.{collect,emit} |
Not a dataplane category; see Event export |
Rhai user metrics
| Pattern |
Controlled by |
conduit_user_<name> |
Script registration + metrics.user_metrics[] collect/emit (defaults follow standard-tier semantics) |
See User metrics.