Send alerts to your team
Every channel takes the same three steps:
- Connect the service on Connections (Owners and Admins).
- Add a channel on Notifications → New channel: the connection plus where exactly to send (a chat, a Slack channel, recipients).
- Route alerts to it on Notifications → Routes, then Save routes.
Below is step 1 and 2 for each service. Step 3 is at the end. Every field is in Connection types.
Telegram
Section titled “Telegram”- Talk to @BotFather, create a bot, and copy its token.
- Add the bot to the group or channel that should get alerts.
- Find the chat id. Group ids start with
-100; a public channel can be written@channelname. - Connections → Telegram → Connect: a name, the Bot token, and optionally a Default chat id.
- New channel: choose the connection, name it, and set Chat id (empty: the connection’s default).
A wrong chat id, or a bot that isn’t in the chat, fails at once with Telegram’s reason, shown under Notifications on the alert.
Choose how to connect (Connect with):
- Incoming webhook: one channel, nothing to install. In Slack, create an app with an incoming webhook for the channel and copy its URL (
https://hooks.slack.com/…). - Bot: any channel the bot is invited to. Create a Slack app with the
chat:writescope, install it, and copy the bot token (xoxb-…). Optionally set a Default channel.
Then New channel. With a bot connection, set Channel (#alerts or a channel id); invite the bot to that channel first, or delivery fails with “Invite the bot to …”.
- Connections → Email (SMTP): your SMTP server or provider (SES, SendGrid, Postmark, Microsoft 365): SMTP host, Port (587 with STARTTLS, or 465 with TLS from the start), username, password, and a From address such as
SREInsights <alerts@example.com>. - New channel: Recipients, comma-separated, such as
oncall@example.com, sre@example.com.
Each notice is one email; all emails of one alert thread together.
AWS SNS
Section titled “AWS SNS”An SNS topic passes alerts to its subscribers: email, SMS, a queue, a Lambda function.
- Create the topic in AWS and add subscribers.
- Connections → AWS SNS topic: the Topic ARN and its Region. There are no secrets: SREInsights publishes with its own AWS credentials.
- The instance’s role needs
sns:Publishon the topic. On a serverless instance, list the topic insns_topic_arns(AWS serverless). On the shared hosted product, ask the operator. - New channel with the connection.
Subjects are plain ASCII of at most 100 characters, such as [high] checkout API: Check failing. A FIFO topic gets the alert as its message group, so one alert’s notices stay in order.
PagerDuty
Section titled “PagerDuty”- In PagerDuty, add an Events API v2 integration to the service and copy its integration key.
- Connections → PagerDuty: the key, and the region (
usoreu). - New channel with the connection.
Each alert is one incident: a re-fire or escalation updates it, an acknowledgement acknowledges it, and a resolve resolves it.
Others
Section titled “Others”- Jira Cloud: one issue per alert, with comments when it fires again and an optional transition on resolve. The channel names the Project key.
- Microsoft Teams, Discord, Google Chat: a webhook URL from the channel or space.
- Webhook: your own endpoint receives each notice as JSON, signed with
X-SREInsights-Signaturewhen you set a signing secret.
Route alerts to the channel
Section titled “Route alerts to the channel”On Notifications, under Routes:
- Add route.
- Severity: leave empty for all, or pick some, such as
criticalandhigh. - Labels: leave empty, or match labels such as
team=payments. - Send to: your channel.
- Tick Keep matching routes below this one if later routes should also get the alert.
- Try an alert to check, then Save routes.
Routes are checked top to bottom, and the first match decides unless it keeps matching. An alert no route matches is recorded but notifies nobody. More in Channels and routes.
Test it
Section titled “Test it”Make a monitor fail on purpose (point it at a page that doesn’t exist) and select Run now. The alert’s Notifications section shows what each channel was sent and whether it went through. Failed sends are retried for a while with back-off; bad credentials fail at once with the reason.