Alerts, Problems view and notification channels
A single list of what is wrong, sorted by priority, and notifications that reach the place where the team works.
The Problems view
At the top of the sidebar, Problems lists every monitor group of the workspace that is not OK, sorted by priority then age. A count of unacknowledged problems stays visible in the sidebar.
For each line: priority, state, monitor, group, host, value, running duration, acknowledgement, last notification. Actions:
- Acknowledge, with a comment: renotifications and escalations stop; a return to OK clears the acknowledgement. Multiple selection acknowledges in bulk.
- Explain: S33D's AI describes the probable cause (see Explain and repair).
- Put in downtime (see Downtimes).
- Open an incident, pre-filled with the monitor, the group and their history (see Incidents).
Filters: priority, state, acknowledged or not, held by a downtime or a dependency, free text.
Priorities
| Priority | Colour | Suggested use |
|---|---|---|
| P1 | red | service down for customers |
| P2 | orange | outage getting worse or hitting a feature |
| P3 | yellow | degradation to handle within the day |
| P4 | blue | to schedule |
| P5 | grey | information |
Notification channels
Settings › Notification channels (administrator role). Each channel has a Send a test button.
| Channel | Settings |
|---|---|
| one or more addresses; plain HTML message with the state, value, threshold, link to the monitor and to the explanation, and the host's latest log lines | |
| Push notifications | all members of the workspace or selected members; notification on the phones and computers they enabled, with the link to the problem and its explanation, and an Acknowledge button (see Mobile app) |
| Webhook | address and headers; JSON body {monitor, group, state, value, message, url} |
| Slack, Microsoft Teams, Google Chat, Discord, Mattermost | incoming webhook address |
| Telegram | bot token and chat identifier |
| ntfy | server, topic, optional token |
| PagerDuty | routing key (Events API v2): trigger then resolve |
| Opsgenie | API key and region (EU or US): create then close |
| GLPI | REST API address, application token and a user's API token: ticket opened on alert, solved on recovery (see below) |
Each send has 10 seconds to succeed, and two retries. A channel's secrets (tokens, keys, headers) are never shown again.
SMS notifications, planned for the Pro and Entreprise plans (€0.10 excl. VAT each), are not open yet.
GLPI channel
An alert opens a ticket in GLPI through its REST API (apirest.php); the recovery solves it.
- Alert (ALERT, WARN, no data): a ticket is opened, titled with the monitor and group followed by a stable reference (
[HYPHA-1a2b3c4d5e]); its urgency follows the monitor priority (P1 very high to P5 very low). While that ticket is not solved, reminders, escalations and worsening add a follow-up to it instead of opening another ticket. - Recovery: a “recovered” follow-up, then a solution: the ticket becomes “Solved” (closing stays with GLPI and its approval rule).
- Send a test: a ticket opened then solved right away.
Settings: the address (https://your-glpi/apirest.php, or the GLPI address, completed automatically), the application token (Setup › General › API: enable the REST API, then an API client with its token) and a user's API token (Preferences › Remote access keys), preferably a service account with the Technician profile (allowed to create tickets and add follow-ups and solutions). Optional: type (incident or request), entity, ITIL category, request source (GLPI numeric IDs). The API is reached with tokens only: no login or password. A session is opened then closed for each send. A GLPI on a private network is only reached when the workspace allows private targets (self-hosting).
Tested against GLPI 12.0.2 (official image, MariaDB 11.4): ticket opened with “very high” urgency for a P1 monitor, follow-up added by the reminder, follow-up and solution on recovery (status “Solved”), channel test, rejected token (the ERROR_GLPI_LOGIN_USER_TOKEN error shown as is).
Escalations
In a monitor's options: if a group stays non-OK and unacknowledged after a delay (for example 15 minutes), HYPHA notifies other channels, with their own message. Several steps can follow each other.
Scheduled reports
Settings › Reports sends by email, on the chosen schedule:
- a summary: host availability, alerts by priority, noisiest monitors, problem duration, SLOs, incidents, hosts that appeared or disappeared, pending security updates, certificates to renew within 30 days;
- the SLO status;
- a dashboard rendered as values and small charts.
Each report has a preview and a send-now button.
Updated: 21 November 2026