Configuration

All runtime configuration lives behind the Settings page (/settings). It is a key-value store that holds strings and returns typed values with a safe fallback, so a bad value never crashes a queue. Every key ships with a default.

Schedule

Scheduling is data, not config keys: each schedule row owns its own escalation. The scheduler is a thin per-minute dispatcher that enqueues a speed-test job according to the configured cron. Edit the cron expression to change how often tests run.

Servers

• Preferred servers — pin tests to specific Ookla servers.
• Blocked servers — exclude servers from selection.

Both accept server-id lists and are resolved through a single reader.

Retention

Set how long measurements are kept. The CleanupWorker prunes old rows on a schedule according to the retention policy.

Degradation thresholds

Define what counts as “healthy.” Thresholds resolve as per-row override → global Settings fallback, with one reader (thresholds_for/1) so there is a single source of truth and resolution never yields nil. Register the global default in Settings.@default_settings.

Notification channels

Notifications are behaviour implementations, not branches in a worker. The worker orchestrates; the channel behaviour does the work. Add a new channel by implementing the behaviour.

Ping target

A schedule may override the ping target per row; otherwise it falls back to the global setting:

schedule.target_host || Settings.get("ping_target")