Skip to content

When Hora fits

Hora is built around one promise: it wakes you only for a real outage. Everything else follows from that: a failed check is retried, an outage needs several failures in a row, peers in other places can confirm it, and a cascade sends one alert about its cause. It ships as one small binary with a TOML file, so the whole setup fits in git.

That focus makes it a good fit for some teams and not for others. This page tries to help you tell which one you are.

  • Your nights matter more than your dashboards. You want few alerts, and every one of them worth acting on: retries, fail_threshold, maintenance windows and silences are there to keep the pager quiet until it should not be.
  • You run services that depend on each other. With depends_on, a down database is one alert about the database, not one per service behind it (see Concepts).
  • You can watch from more than one place. Two or three small Hora nodes watch each other and confirm an outage before anyone is paged; a hiccup seen from one place only is reported as such.
  • You keep your configuration in git. One TOML file, reloaded live, the same on every node, reviewed in pull requests like the rest of your infrastructure. No web editor to drift from it.
  • You make promises about uptime. Error budgets, burn-rate alerts and a printable monthly report come with every monitor that sets a target (SLOs).
  • You want it small. One static binary, SQLite, a calm public page that stays fast with hundreds of services and months of history.

Self-hosted monitoring has good neighbours, and they are worth knowing.

  • You would rather click than write a file, or you need a check type Hora does not have (databases, MQTT, a real browser…). Uptime Kuma has a friendly web UI and a very long list of monitor types and notification services. Coming from it? hora import kuma converts a backup.
  • You need UDP, SSH or gRPC checks, or multi-step check suites, in a single binary and a YAML file. Gatus is close to Hora in spirit and covers those.
  • What you watch is mostly scheduled jobs. Healthchecks does that one thing very well, with time zones and systemd OnCalendar schedules. Hora’s push monitors cover the common case next to active checks; the two also combine nicely: a Hora node can send its own dead-man heartbeat to a Healthchecks check, so something outside watches the watcher.
  • Some check types. No built-in UDP, SSH, gRPC, database or browser checks (see the roadmap). An exec probe runs any monitoring-plugins check, which covers many of them.
  • A long list of notification services. Ten channels are built in; the generic webhook reaches the rest.
  • A web editor. Everything lives in the TOML file, by design.
  • Multi-step scenarios (log in, then check), to keep the configuration simple.
  • A database server. SQLite only: one file, easy to back up.