Skip to content

Operator metrics bases

metrics.base sets how rich Conduit's built-in metrics are when you scrape the Prometheus metrics endpoint. A value of minimal keeps query labels coarse and skips timing histograms which constitutes metrics with lower metric cardinality. standard adds fine labels (qtype, qclass, ip_family), phase and forward timing series, and process gauges — more detail, more series.

Operators usually need to answer: “Is minimal enough for my dashboards, or do I need standard?” This walkthrough runs the same DNS traffic under each base, scrapes /metrics, and shows the concrete differences so you can choose a base (or confirm a reload changed cardinality as you expected).

Prerequisites: Conduit built; upstream DNS resolver listening on 127.0.0.1:5300; Prometheus scrape available via curl (no Prometheus server required for this lab). Enable pools[].health if you want to confirm health series on minimal.

What you will practice

  1. Start Conduit with base: minimal, send a few queries, and read the scrape
  2. Switch to base: standard (restart or overlay) and repeat under the same traffic pattern
  3. Compare label sets on conduit_queries_total and whether phase/forward timing series appear

What you will see

Base Hot-path emphasis After a few A queries you should see…
minimal Volume + failures + lookup + health + topology + meta conduit_queries_total with listener + protocol only; failure counters when applicable; no conduit_phase_duration_seconds; health gauges when probes run
standard Rich labels + timing + process conduit_queries_total with qtype, qclass, ip_family; phase and forward histograms/counters

1. Minimal base

Save as conduit-metrics-minimal.yaml:

schema_version: 1
listeners:
  listeners:
    - address: "127.0.0.1:15353"
      protocol: udp
pools:
  - name: default
    backends:
      - address: "127.0.0.1:5300"
    health:
      enabled: true
      interval_ms: 1000
metrics:
  enabled: true
  base: minimal
  prometheus:
    listen_address: "127.0.0.1:9090"
    path: /metrics

conduitctl validate --file conduit-metrics-minimal.yaml
conduit /path/to/conduit-metrics-minimal.yaml

Send a few queries:

dig @127.0.0.1 -p 15353 +time=3 profiles-minimal.example.com A
dig @127.0.0.1 -p 15353 +time=3 profiles-minimal.example.com AAAA

Scrape and inspect query counters:

curl -sS "http://127.0.0.1:9090/metrics" | grep '^conduit_queries_total'

Expect lines with labels like listener="127.0.0.1:15353" and protocol="udp" — without qtype="A" on the same metric (minimal coarse labels).

Confirm phase histograms are absent:

curl -sS "http://127.0.0.1:9090/metrics" | grep conduit_phase_duration || echo "no phase histograms (expected for minimal)"

After ~1 health interval, confirm health series:

curl -sS "http://127.0.0.1:9090/metrics" | grep -E 'conduit_backend_health_observed|conduit_probe_results_total' | head

Stop Conduit before the next step (or use overlay apply if the control plane is enabled — see below).

2. Standard base

Copy the config to conduit-metrics-standard.yaml and change the base:

metrics:
  enabled: true
  base: standard
  prometheus:
    listen_address: "127.0.0.1:9090"
    path: /metrics

With a control plane enabled, you can apply an overlay that only changes metrics.base without restart. Otherwise restart Conduit with the new file.

dig @127.0.0.1 -p 15353 +time=3 profiles-standard.example.com A
curl -sS "http://127.0.0.1:9090/metrics" | grep '^conduit_queries_total'
curl -sS "http://127.0.0.1:9090/metrics" | grep conduit_phase_duration | head

Expect qtype/qclass/ip_family on queries and phase/forward timing series present.

Legacy profile alias

metrics.profile: full still loads and maps to base: standard (with a deprecation warning). Prefer base: in new configs.