rajnandan1/kener v4.1.0

Release of v4.1.0

on GitHub

New features {#new-features}

Confirmation Threshold (grace period) {#confirmation-threshold}

Monitors can now require N consecutive checks before a status change is committed, so transient blips no longer dent uptime % or the day bars.

• Set a per-monitor Grace period (1–60, default 1 = off) in Monitor → General Settings, or via confirmation_threshold on the v4 monitor create/update API.
• Damping happens once in the scheduled-check write path. A new monitoring_data.raw_status column records what each check actually observed, while status holds the confirmed side.
• On the Nth consecutive opposite observation the status flips and the pending window is retroactively backfilled — a confirmed outage reads as down from its first failing check, and recovery is symmetric.
• Everything downstream — uptime %, day bars, badges, alert evaluation, group scoring — inherits the damped status automatically. Alerts compose by max(grace period, failure threshold).

See Grace Period for full details.

Manual maintenance event transitions {#maintenance-manual-transitions}

Individual maintenance events can now be manually completed or cancelled from the maintenance management UI, letting the recorded timeline reflect what actually happened.

• New Complete and Cancel actions move an event to a terminal COMPLETED or CANCELLED status and adjust end_date_time accordingly.
• Terminal statuses are final — a completed or cancelled event can't be edited further.
• The v4 API supports status transitions alongside window edits, with the two enforced as mutually exclusive.
• Subscribers are notified of the transition.

See Maintenance Events.

Last known status for "None" monitors {#none-last-known-status}

The None monitor type can now optionally reuse the last known status (status, latency, and source) instead of reporting no data when no value is pushed. Enable it with the new checkbox in the monitor configuration UI.

Page settings {#page-settings}

Status pages gained per-page presentation controls, configurable in the manage UI and over the v4 pages API:

• Per-page status history — separate day counts for mobile (default 30) and desktop (default 90), each bounded to 1–365.
• Monitor layout styles — choose between default-list, default-grid, compact-list, and compact-grid.
• Meta / social overrides — per-page title, description, and preview image.

Page settings accept deep-merge PATCH payloads over the API, and invalid payloads are now rejected with validation errors.

Inline monitor toggles {#inline-monitor-toggles}

The Manage → Monitors list now lets you flip a monitor's active/inactive state and visibility inline, without opening the full edit page for a one-bit change.

API: address the home page with ~home {#api-home-token}

The home page is stored with an empty path and previously couldn't be addressed over the REST API. GET/PATCH/DELETE /api/v4/pages/~home now targets it, and list and single-page responses render its page_path as ~home so what you read is exactly what you can address. PATCH accepts every field except the (fixed) home page path, and DELETE is rejected with a 400. Unmatched /api/ routes now return JSON 404s instead of falling through to an HTML page.

API: absolute url on incident and maintenance responses {#api-absolute-url}

v4 API responses now carry an absolute url field so consumers link to the right page instead of concatenating ids onto paths:

The URL is built from the configured Site URL plus base path.

API: delete a monitor by tag {#api-delete-monitor}

A new DELETE /api/v4/monitors/{monitor_tag} endpoint removes a monitor through the same path the manage UI uses. Monitor deletion now also explicitly cleans up alert-config rows (v2 alerts and the trigger/monitor junctions), fixing orphaned rows left behind on SQLite — which never enforced the ON DELETE CASCADE foreign keys. Shared alert configs survive until their last monitor is detached.

Improvements {#improvements}

Rich maintenance descriptions on list views {#maintenance-html-descriptions}

Maintenance descriptions now render as sanitized HTML on list views — the home page, custom pages, monitor pages, events pages, and embeds — matching the detail page. HTML formatting and line breaks display correctly instead of showing literal tags, with sanitization kept since these render to anonymous visitors.

Searchable monitor picker for groups {#group-monitor-search}

The group monitor picker now has a search bar, making it practical to find and add monitors on instances with hundreds of them instead of scrolling a flat list.

Database connection pool tuning and health checks {#db-pool-tuning}

The database layer now tunes connection-pool sizing, clamps the pool bounds to safe values, guards the Redis probe, and hardens the error page. The configured database type is logged at startup for clearer diagnostics.

Broader alert evaluation {#alert-evaluation}

Consecutive status and latency checks now count all alert-visible sample types, improving alert reliability, and manually submitted data triggers a best-effort alert re-evaluation after updates.

Localization {#localization}

• Added a Ukrainian translation.
• Fixed syntax errors in the French locale strings.
• Updated the default v4 alert body description, which still referenced pre-v4 template variables.

Bug fixes {#bug-fixes}

• Overall status now follows worst-wins ordering (DOWN > DEGRADED > MAINTENANCE > UP) consistently across components, so an active maintenance no longer masks a real outage in the aggregated status.
• Social preview images use absolute URLs — og:image and twitter:image meta tags now prepend the Site URL, fixing previews that failed because of relative paths.
• Fixed overlapping header sections on the public page by adding a single frosted backdrop behind the fixed nav and the sticky theme bar, so page content no longer shows through while scrolling.