Monitoring

A service can be checked by Statup itself, every minute, from the server it runs on, so internal addresses work.

Set a check

In the service’s form, under Monitoring, choose:

• None, the default: no check.
• Web, for an address that should answer. URL is a full http:// or https:// address, such as https://intranet.example.com, of up to 2,000 characters. Tick Accept an internal certificate for a tool whose certificate no public authority signed.
• Port, for a host and its port, such as 192.168.1.1 and 443: a server, or a router or a firewall by its administration port. Port is a number from 1 to 65535.

An address that holds a user name or a password is refused: the checks keep no credentials.

Test runs the check at once, without saving, and says why it fails: the port is closed, there is no answer within 10 seconds, the name is unknown, the certificate is refused, the server answers with an error, the address redirects in a loop, or the connection fails.

How a check runs

• Every minute, up to 16 checks at a time, each given 10 seconds.
• Web sends a GET with the user agent statup and follows up to 5 redirects. Any status below 500 counts as an answer, even a refusal such as 401 or 404.
• Port opens a TCP connection and sends nothing. An accepted connection is an answer.
• The checks only reach the addresses you set.

From failures to an outage

When a service fails as many checks in a row as its delay, three by default, it shows Outage, marked Detected automatically, dated from its first failed check. The first check that answers ends it, and the state set by hand comes back.

Each service sets its delay under Monitoring: “Declare an outage after 3 min without an answer”, from 1 to 60. A critical server can show its outage after a minute, a printer after fifteen. Changing the delay, or saving the form with the same check, leaves an outage already shown as it is.

Statup never writes an incident itself. Declare an incident, on the banner or on the Services page, opens a new incident with the impact Major, the service and the start already filled in. See Incidents.

When no outage is declared

• During a maintenance under way that names the service, unless the maintenance is marked No downtime. Failures are counted but raise nothing. If the service still fails when the maintenance ends, the outage is declared at the next failed check, dated from then. A check that answers during the maintenance still ends an outage.
• When Statup itself loses its connection. If at least three services that were answering all fail in the same round, the round is ignored, for at most five rounds in a row. Two services failing together are judged as usual.

Alerts

Services that fall in the same minute, such as the services of a server being restarted, are told in one message per channel (“3 services are not answering”, with their names), and their return in one more. A service that an open incident already covers sends no alert. Declaring a maintenance before a planned restart keeps the checks quiet. See Alerts.

In the history

A detected outage of fifteen minutes or more colours its days in the service’s availability history, unless an incident already covers that day. A shorter one leaves the day green. See Services.

Changing the kind of check, the address, or Accept an internal certificate clears an outage under way, such as a false one caused by a wrong address. Outages already ended stay in the history.

Turn the checks off

MONITORING=false stops the checks that run every minute. The addresses stay saved, and Test still checks an address when you press it.