Monitoring alerts

An alert watches one thing on an instance and emails you when it goes past the limit you set. It emails you again when it clears. An alert never changes anything on the instance. Alerts cost nothing.

Open the instance, go to its Alerts tab and choose Add alert…. Choose what to watch:

For a CPU or down alert, the minutes keep a short spike from setting it off. CPU waits 5 minutes unless you choose otherwise, and down waits 3.

A down alert can also name a port. We then try to connect to that TCP port on the instance's IPv4 address every minute, and call the instance down when it does not answer. Use the port your service listens on, such as 443 for a website or 22 for SSH. Your firewall must let connections to it through.

An instance holds up to 10 alerts.

• Disk use reads the filesystems through the QEMU guest agent. Our images have it. On an image of your own, install and start qemu-guest-agent, or the alert shows Unknown with the reason.
• Down knows when you stopped the instance yourself, from the dashboard or the API, and does not go off then.
• Data transfer this month goes off once a month at most, and clears on the first of the next month.

Each alert shows where it stands now:

• OK: within its limit.
• Over the limit: past its limit, and waiting out its minutes before it goes off.
• Firing: it has gone off and we emailed you.
• Unknown: we could not read the value, with the reason. An unknown alert never goes off. When we cannot reach the host the instance runs on, a down alert is unknown, not down.
• Paused: you paused it.

Its last reading and when we took it are beside it. The History below lists when each alert went off and cleared, and whether we sent the email.

From an alert's menu you can pause it, stop or start its emails, or delete it. Deleting an alert keeps its history.

curl -X POST https://api.coritan.com/api/v1/client/vps/$INSTANCE_UUID/alerts \
  -H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
  -d '{"metric": "cpu", "threshold": 90, "duration_minutes": 10}'

metric is cpu, disk, bandwidth or down. threshold is a percent for cpu and disk (1 to 100) and GB for bandwidth; down takes none. duration_minutes (1 to 1440) applies to cpu and down, and port only to down. A field out of range answers 422 and says which, and an eleventh alert answers 409.

GET /api/v1/client/vps/{uuid}/alerts answers the alerts with their state, last_value, last_detail and last_checked_at, the last 50 events, and the metrics with their ranges. PATCH .../alerts/{rule_uuid} takes threshold, duration_minutes, port, enabled and notify_email. A changed limit starts the alert over at OK.

• Alerts by text message, chat or webhook. We send email only.
• Memory use. The host cannot see how much memory the instance's own system uses.
• Checks from more than one place. We check from Coritan's network.