Observability

Conduit exposes metrics, tracing, event export, and process logging so you can observe the dataplane without changing query behavior. Observation runs off the DNS query path: the export backlog or full observability queues drop data rather than delaying client responses.

For further information

You need…	Page	Config block
Aggregate QPS, pool mix, forward errors, backend health gauges, or SLO dashboards	Metrics	`metrics:`
To tune which built-in series are recorded and exported	Metrics configurability	`metrics:`
Which categories a metric belongs to, and dimension vocabularies	Built-in metric registry	(same `metrics:`)
Names, labels, and PromQL examples for every built-in series	Built-in metrics	(same `metrics:`)
Per-query wire copies to a log analytics or tap collector	Event export	`events:`
Per-query pipeline phases (lookup, forward internals, retries) on selected queries	Tracing	`tracing:`
Process logs for startup, reload, control RPC access, or optional query summaries	Logging	`logging:`

flowchart TD
  Q[What do you need to see?]
  Q -->|Volume and latency over time| M[Metrics]
  Q -->|Full DNS wire per query| E[Event export / dnstap]
  Q -->|Phase timing on one query| T[Tracing + GetTrace]
  Q -->|Process and config events| L[Logging]
  M --> BM[Built-in metrics catalog]

For a lab walkthrough that enables metrics and tracing together, see Metrics and tracing. For OTLP HTTP metrics push with conduit-otlp-metrics-tracer, see OTLP metrics push smoke. For dnstap export with conduit-dnstap-tracer, see Event export and dnstap. Config fields: Reference: metrics and tracing, Reference: events.

OTEL and tracing names

Conduit uses several words that sound similar but mean different things:

Term	What it is	Config
OTLP metrics	Periodic push of built-in counters/gauges/histograms to a collector	`metrics.otel`
Pipeline trace	In-memory per-query phase timeline (`GetTrace`, `conduitctl trace`)	`tracing:`
Process logging	Rust `tracing` subscriber to stderr/stdout	`logging:`
`logging.level: trace`	Maximum log verbosity — not pipeline tracing	`logging.level`

Not implemented today: OTLP distributed traces (/v1/traces) and OTLP logs (/v1/logs) export. See Metrics, Tracing, and Logging for what ships now.

Correlating per-query signals

When you debug a single query across tracing, logs, and dnstap, use the internal transaction id (txn_id):

Signal	Where `txn_id` appears
Logging	`query complete` / `query dropped` lines at `debug` (`txn_id=…`)
Tracing	Argument to `conduitctl trace` and gRPC `GetTrace`
Event export	Optional `extra_fields: txn_id` on dnstap frames

Enable logging.level: debug briefly to read txn_id from a completed query, then fetch the pipeline trace for that id. Traces expire after 5 minutes or when the store exceeds 1000 entries — see Tracing.

Built-in metrics never label series with qname, client IP, or txn_id. Use event export or tracing for per-name or per-transaction detail.

Changing observability config

Reload and restart rules differ by block. After editing the file, reload (SIGHUP or conduitctl reload) updates the stored config and snapshot where supported; some changes still need a process restart to take effect on running listeners or export tasks.

Block	Overlay allowed?	Reload updates snapshot?	Restart required for…
`metrics:`	Yes (deep merge)	Yes	Rare cold cases only — plan knobs and Prometheus/OTLP bind apply live; see Metrics configurability — Overlay
`tracing:`	No	Yes (stored config)	Enabling tracing, activation rules, `log_json`
`events:`	Yes	Yes	Adding/removing sinks, destinations, `queue_depth` (filter/emit/extra_fields on existing sinks reload)
`logging:`	Yes	Yes (stored config)	`level` or `output` (subscriber binds at process start)

Details: Configuration model — What takes effect when, Reload and export.

Topic pages

Metrics — enable export, minimal vs full profiles, Prometheus scrape and OTEL push
Built-in metrics — series catalog, profiles comparison, PromQL examples
Tracing — per-query pipeline traces
Event export — dnstap sinks
Logging — process log level, stderr/stdout, lab smoke test

Lab guides: Metrics and tracing · Operator metrics bases · Metrics beyond bases · OTLP metrics push smoke · Event export and dnstap

Symptom-oriented help: Troubleshooting.