Alert Channels

Where Zharp sends alerts when a monitor goes down, degrades or recovers: email, the mobile app, and the chat, paging and automation tools you connect under Integrations & API.

Where alerts go

Every alert integration lives under Integrations & API in the dashboard sidebar. The ones that receive alerts are:

Two more routes don't need an integration:

• Email. When a monitor goes down or comes back up, Zharp emails the workspace owner, everyone subscribed to the monitor, and the subscribers of status pages that show it. The owner and monitor subscribers only get these e-mails if their own settings allow it — see Your alert settings. To send alerts to a shared inbox or a distribution list instead, use an e-mail connection.
• On-call paging. A monitor's How should we escalate incidents? section decides who is paged and how — E-mail, Push notification or Slack alert. See Escalation & On-call.

Connect an integration

1. 1
   Open the integration

   Go to Integrations & API and pick the platform. Its page lists the connections you already have.

2. 2
   Add a connection

   Click Add Connection, paste what the platform gives you (see the table below), and choose the events under Trigger on. Click Add connection to save.

3. 3
   Choose its monitors

   The new connection opens on its own page. Set Applies to to All monitors or Selected monitors, then click Send test alert to check it.

Slack

Add Connection takes you to Slack's own consent screen, where you pick the workspace and channel. When you come back, the connection is already set to All monitors with Down, Degraded and Resolved on.

Telegram

Telegram connects through the Zharp bot — there is no bot token or chat ID to copy. Add Connection opens Connect Telegram with two tabs:

• Direct message — click Open in Telegram (or scan the QR code on your phone) and press Start.
• Group / channel — add the bot to the group, then send the /start command shown in the modal, code included. For a channel, add the bot as an admin with the “Post messages” permission and post the command there.

The modal waits and links the chat as soon as the bot hears from it. If the code expires, click Get a new one. Like Slack, a new Telegram connection covers All monitors with Down, Degraded and Resolved on.

Zapier

Open Zapier takes you to Zapier to build a Zap. A Zap built with the Zharp app in Zapier shows up under Your Zaps. To use a Catch Hook instead, click Add a Catch Hook URL and paste the Zapier webhook URL (https://hooks.zapier.com/…). The Zapier guide covers the triggers, what each alert contains, and the key scope the app needs.

E-mail connections

An e-mail connection sends alerts to fixed addresses — a team inbox or a mailing list — rather than to people in the workspace. It has no page under Integrations & API yet: create it through the API with an admin key, listing the addresses in config.emails. A connection without at least one valid address is refused.

curl -X POST https://api.zharp.io/api/v1/notification-channels \
  -H "Authorization: Bearer $ZHARP_ADMIN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "workspaceId": 12,
    "type": "EMAIL",
    "name": "Ops inbox",
    "config": { "emails": ["ops@example.com", "oncall@example.com"] }
  }'

After that it behaves like any other connection: PUT /notification-channels/:id/monitors chooses its monitors, its Trigger on events, delay and repeats apply, escalation steps can name it, and POST /notification-channels/:id/test sends each address a test e-mail. Every alert is one e-mail per address, linking to the incident in the dashboard.

Events

Each connection chooses its events under Trigger on when you add it, and under Alerts routed here on its page afterwards (Events sent for a webhook). New connections start with Down, Degraded and Resolved.

PagerDuty

PagerDuty offers two choices, named by severity: Critical (Down) and High (Degraded). Down is sent with PagerDuty severity critical; Degraded with error — PagerDuty has no “high” severity, and error is the level its severity-based urgency treats as high urgency. Both use the same alert per monitor. Resolved and Acknowledged are always sent, so the alert in PagerDuty is acknowledged and closed along with the incident. Maintenance is never sent to PagerDuty.

Choose which monitors alert

A connection only hears about the monitors it covers. Set this under Applies to on the connection's page:

• All monitors — every monitor, including ones added later.
• Selected monitors — only the monitors you tick in the list below it.

For Slack you can also pick connections from the monitor itself: tick Slack alert under How should we notify the on-duty member? and choose them in Send Slack alerts through.

An escalation policy step can name a connection too. That step posts the incident whatever the connection's Trigger on says — the policy chose it. See Escalation & On-call. So can Report to when you report an incident by hand.

Pause a connection

Click Pause at the top of a connection's page to stop everything going to it — alerts, repeats, escalation steps and reports — without disconnecting it. The connection shows Paused until you click Resume. Send test alert still works while it's paused.

Test messages and retries

Send test alert (Send test event for a webhook) on a connection's page sends a message saying the connection is set up correctly. PagerDuty gets a low-severity test alert that is resolved straight away. Every delivery from the last 24 hours shows under Recent deliveries, and the Status card shows whether the last one got through.

To test from a monitor instead, open it and click Test Notification (or Send test alert in the monitor list's menu). Every connection that alerts for the monitor gets a test alert for [Test] monitor name, saying nothing is wrong with the monitor; with Send to attached notify-only users on, the monitor's subscribers get a test e-mail too. Nobody is paged and no incident is opened. A message then says where it went, and names any connection that was skipped (a paused one, for example) or couldn't be queued.

Failed deliveries are retried with exponential backoff, starting 5 seconds after the first try. A connection makes 3 attempts by default. On a webhook endpoint's page you can change this under Retries to 1, 3 or 5 attempts — see Webhooks for timeouts and response codes.

Delay and repeat

Each link between a monitor and a connection has its own timing. Set it on the monitor: Edit monitor → Integrations & team → Connection alerts, which lists every connection linked to the monitor with three choices:

A repeat goes only to the connection that set it. It doesn't e-mail or push anyone, and it never pages on-call again — repeat paging belongs to escalation policies. Each repeat is noted on the incident's timeline (“Still down — repeated the alert to #ops (2 of 3)”). Paused connections and monitors with alerts muted get no repeats.

Through the API, the same settings sit in the monitor's notificationChannels list:

curl -X PUT https://api.zharp.io/api/v1/monitors/3381 \
  -H "Authorization: Bearer $ZHARP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "notificationChannels": [
      { "channelId": 41, "delay": 120, "repeatInterval": 30, "repeat": 3 }
    ]
  }'

Your alert settings

What reaches you personally is set under Settings → Alerts & Reporting, or Notifications in the mobile app — the same settings in both places:

Mobile app push alerts

Install Zharp from Google Play and sign in — the device registers itself (the iPhone app is on its way: join the waitlist). It then shows under Integrations & API → iOS or Android, in Registered devices.

• When a monitor goes down or comes back up, the workspace owner and the monitor's subscribers get a push if Push Notifications is on under Settings → Alerts & Reporting. Its event menu chooses Down Events and Up Events, and SSL Expiry / Domain Expiry for SSL and domain monitors.
• On-call pages arrive as a push when the escalation asks for one.
• Open a device to rename it or click Pause alerts. A paused device gets nothing until you resume it. Only the device's owner can change it.