Channels and delivery

What it is#

Alerting (sidebar) is one page for what is wrong and who gets told. Its Alarms tabs (Open now, History, Rules, Maintenance) are described in Alarms and rules. This chapter covers the Delivery tabs and the help tab:

Where to find it#

In the sidebar, open Alerting, then Channels, Delivery log or Muted. Direct links: /dashboard/alarms?tab=channels, ?tab=deliveries, ?tab=muted, ?tab=how. The old /dashboard/notifications address still works and opens the same tab. Add channel is in the page header on every tab.

The figures at the top#

When there is no channel at all, a banner warns you: incidents still open, but nobody is told.

Adding a channel#

Click Add channel, then:

1. Platform. Each platform tile lists the steps to get its webhook:

   Platform	You enter
   Slack	The Incoming Webhook URL; optionally a channel override (only older webhooks use it)
   Discord	The channel webhook URL
   Microsoft Teams	The URL from a Teams workflow or connector
   Google Chat	The space webhook URL
   Email	One or more addresses, separated by commas (at most 20)
   Webhook	URL, method (POST, PUT or PATCH) and optional headers, e.g. Authorization

2. Name. Something people recognize, e.g. "Slack – production on-call".

3. Severity. Everything (info, warning, critical and the digest), Warning and up, or Critical only (a pager or on-call channel). See Alarms and rules.

4. Projects (when there is more than one). Only alarms about the chosen projects; nothing chosen means every project. Alarms that belong to no project always come.

5. Repeat while firing. While a critical alarm stays unacknowledged, this channel gets a reminder every 5 min, 15 min, 30 min, 1 h, 2 h, 4 h, or a custom number of minutes (1–1440).

6. Which alerts. Choose Everything, or only alerts about Applications, Databases, Hosts or Containers. The category comes from the endpoint's linked metrics. Alerts without linked metrics always go to every channel.

7. Send a test after saving (on by default for a new channel). The test goes through the same path as real alerts, and the result shows when the channel is saved.

The form scrolls on small screens, and the buttons stay at the bottom. If something is missing (a name, a URL that doesn't start with https://), the reason shows next to the button.

Secrets#

Webhook URLs and header values are secrets, so the page never shows them in full. You see https://hooks.slack.com/…a1b2. When you edit a channel:

• leave the URL field empty to keep the saved one;
• for headers, leave the value empty to keep it;
• type a new value to replace it.

If you change a channel's platform, you need to enter a new URL.

The channel card#

The actions are:

• Send test: sends a real message now and shows Test delivered in 0.8 s or the platform's error.
• Log: opens the delivery log for this channel.
• Pause / Resume: a paused channel gets nothing, and keeps its settings.
• Edit.
• Delete: administrators only. It also removes the channel's delivery history and its mutes. The dialog offers Pause instead.

Members and administrators can add, edit, test and pause channels; viewers can only look.

Delivery log#

The log shows every attempt, newest first:

• a coloured dot for the kind of alert: red for down or escalated, green for recovered, amber for warnings, blue for tests;
• the subject, the channel, the alert kind and the endpoint;
• the result: Sent, Failed or Pending;
• when it happened.

Click an entry to see the full message, the platform's error if any, when it was queued and delivered, and a link to the endpoint.

You can filter by channel and by result, hide tests, and load older entries 50 at a time. The log refreshes every 15 seconds.

Common errors:

Muted#

This tab lists every endpoint whose alerts are switched off. For each one you see whether all channels or just one are muted, the reason, who muted it, and until when ("until unmuted" if it has no end).

• Unmute turns alerts back on (members and administrators).
• To mute an endpoint, open its card on Services → Endpoints.

How alerts are sent#

Every alert is an alarm with a lifecycle (Alarms and rules): it is sent when it opens and when it turns critical, reminded while critical and unacknowledged, held during maintenance or while an upstream service is down, and closed with a Resolved message to the channels that heard about it.

What each platform receives#

The webhook JSON looks like this:

{
  "event": "endpoint_down",
  "label": "Down",
  "subject": "🚨 Service Down: Orders API",
  "message": "…",
  "endpoint_id": "…",
  "channel": "PagerDuty bridge",
  "timestamp": "2026-10-14T09:12:03Z",
  "source": "pulsimo",
  "severity": "critical",
  "alarm": {
    "id": "…", "kind": "endpoint_down", "source": "check", "severity": "critical",
    "state": "firing", "value": null, "threshold": null, "unit": null,
    "first_fired_at": "2026-10-14T09:12:01Z", "incident_id": "…", "project_id": "…"
  }
}

alarm is there for every alarm message (not for tests and digests). Deduplicate on alarm.id; a message with alarm.state resolved closes it.

event is one of:

• endpoint_down
• endpoint_recovered
• endpoint_partial_outage
• endpoint_warning
• org_major_outage
• incident_escalated
• ssl_certificate_expiring
• test

Email setup#

Email uses the notification service's environment (.env):

SMTP_HOST=smtp.gmail.com
SMTP_PORT=587
SMTP_USERNAME=alerts@company.com
SMTP_PASSWORD=app-password
FROM_EMAIL=alerts@company.com

After changing them, restart the notification service:

docker compose up -d notification

Then use Send test on an email channel.

API#

Errors come back with a message and a status: 403 when you're not allowed, 404 when the channel doesn't exist, 422 when the settings are invalid, and 504 when the notification service didn't answer the test.

Limits#

• A test or an alert to a platform that is slow to answer can take up to 15 seconds. The test gives up after 20 seconds.
• Delivery history older than 30 days doesn't count in the channel figures. It stays in the log until it's cleaned up by the maintenance retention settings.