Skip to main content

Contact points

A contact point is where notifications are delivered. Open Insights → Alerting → Contact points.

Set one up before you write your first rule. A rule that fires with nowhere to send is silent, and it looks exactly like a rule that is working.

Supported integrations

Each has its own setup guide, covering what to create on the provider's side and what to put in each field.

IntegrationWhat you need
SlackAn incoming webhook URL, or a bot token and a channel
EmailYour own SMTP server, and the addresses to send to
PagerDutyAn Events API v2 integration key
WebhookA URL, and optionally auth
OpsgenieAn API key
Microsoft TeamsA workflow or connector webhook URL
TelegramA bot token and a chat ID
DiscordA webhook URL
Amazon SNSA topic ARN, a region and credentials

A destination without a dedicated integration is usually reachable with Webhook, which posts a JSON body you can parse.

Create one

New contact point, give it a name, pick a type, fill in the fields.

The name is what you pick in notification policies, so name it after the audience rather than the tool: payments-oncall reads better in a routing tree than slack-webhook-2.

Add another integration puts a second destination under the same name. A contact point called payments-oncall can page PagerDuty and post to Slack at once, and the routing tree still only has one thing to point at.

Test it

Test sends a notification through the contact point using a sample alert.

Do this every time you create one. A webhook with a typo in the URL fails silently for weeks otherwise.

Test notifications are real

The test is delivered for real. Anyone watching that channel, inbox or pager is notified. Warn the on-call before testing a paging integration.

Secrets

Tokens, keys and passwords are stored write-only. Once saved, the field reads as set and the value is never sent back to the browser. Admins and owners can reveal a stored secret; editors cannot.

To replace a secret, type the new value over the field. To keep it, leave the field alone.

Changing the destination clears the secret

Change a contact point's type, or the address it sends to, and you have to re-enter its secrets in the same save. A save that changes the destination while leaving the secret field untouched is refused with re-enter the contact point secrets when changing its type or destination.

This stops a stored credential being carried to an address it was not issued for.

What the notification says

Without a template, a notification leads with the first alert's summary annotation, falling back to the rule name, and lists every alert in the group with its labels. Set the summary annotation on your rules and most notifications read well with no template at all.

To change the wording, see Notification templates.

Reachable addresses

A webhook, SMTP host or custom endpoint has to be reachable on the public internet. Delivery to private, loopback, link-local and cloud metadata addresses is refused, and redirects to them are refused too.

Custom headers naming a tenant, such as X-Scope-OrgID, are dropped from outbound webhook requests.

To reach a service inside your own network, put a relay in front of it that has a public address.

Health

Each contact point shows a health chip. A warning means the routing tree does not deliver to it, which usually means a name changed on one side and not the other.

Provisioned contact points

A contact point marked Provisioned is managed outside the portal and reads as read-only. Change it wherever it is defined.

Delete one

Deleting removes every integration under the name, and any notification policy routing there stops delivering. Check Notification policies first.

Permissions

ActionRole
View contact points and which settings are filled inMember
Create, edit, test and delete themEditor
Read a stored secret backAdmin

Next