v0.2.0

Added

• Notification sinks: signed webhook, Slack and Microsoft Teams, with routes that pick
  sinks by severity and event, durable queues, retries, batching and a heartbeat.
  notify-test sends a test event to each sink.
• The monitor UI gains a peer picker, a check builder, a live probe that gates Save,
  and a Routes view with a route preview.
• collector.local_only runs the collector without tailscaled, for the Kubernetes
  ProxyGroup sidecar. The Kubernetes how-to and examples/kubernetes/ show four ways
  to inject it.
• A systemd deployment guide and example unit, a how-to for exposing the UI as a
  Tailscale Service, and a tutorial for receiving a first webhook.

Changed

• HTTP request logs are structured fields instead of a JSON string in msg.
• Audit and request logs record the tailnet caller behind tailscale serve instead
  of the proxy's loopback address.
• The example systemd unit creates the state directory with mode 0700, so the
  collector no longer warns about state_dir on first start.

Full Changelog: v0.1.0...v0.2.0

v0.1.0

First release of tail-watch, a tailnet reachability monitor. It runs beside a
kernel-mode tailscaled, checks your nodes and the subnets behind them, and
tells you which layer failed.

Highlights

• Checks: icmp, tcp, http, and tsping, configured from one YAML
  file. --once runs every check once, and check-config validates a file
  without contacting tailscaled.
• Failure layers: a failed check is labeled unrouted, control,
  tunnel, service, or lan. Remote LAN targets learn a parent from disco
  and primary routes.
• Alert state: targets move between UNKNOWN, UP, DOWN, and
  SUPPRESSED after consecutive failures, with flap damping and suppression
  of dependents behind a down reachability parent.
• Notifications: slack and webhook sinks with grouping, retries, and
  HMAC-signed webhook payloads. Incidents and queues persist in state_dir
  and replay at startup, giving at-least-once delivery. tail-watch notify-test sends a test event to one sink or all of them.
• Heartbeat: an optional dead-man's-switch ping on an interval.
• Monitor UI: an opt-in web UI and JSON API (--ui or collector.ui) to
  add, edit, and delete monitors on a running collector. It has no login, so
  read the risks first.
• Metrics: Prometheus series under the tailwatch_ prefix, served at
  /metrics.

Documentation

Start with the first-monitor tutorial, or
see the documentation index. Each archive includes the
docs/ tree, examples/tail-watch.yaml, README.md, and the license.

Downloads

Archives for Linux, macOS, and Windows on amd64 and arm64, with
checksums.txt. tail-watch --version prints 0.1.0.