Skip to content

Release notes — 1.2.0

Released 2026-07-26 with DNS Conduit 1.2.0.

Packaging

  • Guide lab examples: Production tarballs and the conduit .deb ship primary runnable configs from the Guides under examples/ (package path /usr/share/doc/conduit/examples/), including health, cache, failover, rule action order, metrics labs, dnstap, and the Rhai blocklist tree. See Install and run.

Metrics configurability

  • Bases and categories: Prefer metrics.base (none | minimal | standard) with optional categories.include / exclude. standard is a curated bundle, not the entire registry. A category listed in both include and exclude emits a warning and exclude wins. See Metrics configurability and Built-in metric registry.
  • Collect vs emit: Per-category and per-user-metric collect / emit. Collect-only still costs hot-path recording; it only skips export.
  • Granularity: granularity.default and per-family dimension lists; response rcode coarse vs IANA.
  • Event export axis: metrics.event_export.{collect,emit} for conduit_events_*.
  • Overlay: metrics allowed with deep merge. Plan changes apply live; Prometheus listen rebinds; OTLP reconnects; bind/reconnect failure rejects apply and keeps last-good.
  • Consumer validation: Write sites (metrics.inc / inc_labels) with collect or emit off warn (increments no-op / series stay out of export) but validate/apply succeed — same model as built-in categories. On base: minimal, unlisted script metrics default to collect off; opt them in under user_metrics (or use base: standard) when you want them recorded. Future read APIs will reject collect-off while they still reference a metric. See User metrics.
  • User metric HELP: Optional metrics.user_metrics[].help sets Prometheus # HELP and the OTel instrument description for conduit_user_* (default remains Rhai user-defined metric when omitted).
  • Minimal includes health: base: minimal includes the health category (probe / backend health series when health is configured).
  • Process metrics: process category (standard) exposes Linux scrape series for RSS, open/max FDs, thread count, and CPU seconds (conduit_process_*). See Built-in metrics — Process and build.
  • Uptime: meta includes conduit_uptime_seconds (monotonic seconds since process start, scrape-refreshed). Prefer it over deriving uptime from conduit_start_time_seconds when wall clocks may jump.

Upgrade from metrics.profile

Prefer metrics.base and collect / emit. Mapping for former profile and user_metrics[].export: Metrics configurability — Legacy aliases. Aliases remain accepted through the 1.x line (deprecation warnings); removal is scheduled for the 2.x breaking line.

All changes in this release (automated pull request list).