Hosts and checks
UCP (Uptime Check) checks all the time that your servers, firewalls, sites and services answer, and alerts you when something stops working. A host is a machine to watch, such as a server, a firewall, a printer or a site. A check is a test on that host, such as "the site answers", "port 443 is open" or "the certificate does not expire soon".
Where to find it
- In the subscription, the Hosts tile lists the hosts of that subscription. The number on the tile says how many problems are open.
- In the menu, the Modules group has Hosts, Problems and Downtimes, with the hosts of all your subscriptions.
The plan sets how many hosts and checks the subscription may have, and how often they may run: the page shows the count, for example "4 of 10 hosts" and "12 of 50 checks".
Adding a host
Add host asks for:
| Field | What to write |
|---|---|
| Name | A name for you, such as "Office firewall" |
| Address | An IP address or a DNS name. A name is resolved again at every check |
| Host check | The check that says whether the host is Up or Down: Ping for most hosts; TCP port, with a port that is always open, for a host that does not answer the ping |
| Check every | How often the host check runs. The plan sets the shortest interval |
The first check runs within a minute.
Choose a host check that fails only when the whole host cannot be reached: when the host is down, its other checks do not alert, only the host does.
Adding checks
Open the host and press Add check. Choose the type:
| Type | What it checks |
|---|---|
| Ping | The host answers, within the times and the packet loss you set |
| TCP port | A port accepts connections, and how fast |
| HTTP/HTTPS | A page answers, with the status code and the text you expect |
| DNS | A name resolves, to the value you expect |
| TLS certificate | The certificate is valid and does not expire within the days you set |
Examples
- A site: an HTTP/HTTPS check with HTTPS on, the Name of the site if the host serves more than one, and perhaps a Text in the page that appears only when the site really works. Also check the certificate (with HTTPS) adds a TLS certificate check together with it, which alerts before the certificate expires (it counts as one more check).
- A mail server: a TCP port check on port 25, and a TLS certificate check with SMTP with STARTTLS.
- The dynamic DNS name of a firewall: a DNS check with the Expected value, asking The authoritative servers of the name, which show a change at once. In Addresses of the backup lines write the addresses of a spare line, such as an LTE line: when the name answers one of them, the check goes to Warning instead of Critical.
When a check runs
In When it runs:
- Check every: the normal interval;
- Retry every: after a failure, the check is repeated more often;
- Attempts: how many failures in a row confirm the problem.
With 3 attempts and a retry every minute, a problem is confirmed two minutes after the first failure. So a single lost packet does not wake you up at night. A check that answers again is OK at once.
The states
| State | What it means |
|---|---|
| Up / Down | The host answers or does not answer the host check |
| OK | The check works |
| Warning | The check works badly, for example it is slow or the certificate expires soon |
| Critical | The check does not work |
| Unknown | The check could not be run, for example because the name of the host does not resolve: read the output |
| Pending | Before the first result |
| Not checked | The host or the check is disabled |
"to be confirmed" marks a failure that still has attempts left.
The page of a host
The page of a host shows:
- the State and since when;
- the Checks, each with its state, the Last result and when it ran. Open a check to see its history;
- the Availability of the host: the share of time Up over 24 hours, 7 days, 30 days and 12 months; the bar of The last 90 days, one segment per day (green from 99.9 %, yellow from 99 %, red below, grey without data); the Response time, last 7 days. Downtimes do not count. It is updated every hour;
- the Downtimes (see Downtimes and availability);
- the Changes of state, the most recent first.
Try now runs all the checks at once and shows what they answer, without writing anything in the history: it is useful after changing a parameter or fixing a service.
Edit changes name, address, note and host check. Disable stops the host with all its checks without deleting anything; Enable starts it again. Delete removes the host, its checks and their history.
Problems
Modules › Problems shows what is wrong now:
- Hosts down: the hosts whose host check failed all its attempts;
- Checks: the checks with a confirmed state Warning, Critical or Unknown. A check of a host that is down is marked "host down": look at the host first.
A problem leaves the page when the check is OK again.
Acknowledging a problem
On the page of the check, Acknowledge says that someone is dealing with it, with a Note: the alerts of that problem are no longer repeated, for anyone. The problem stays on the page marked Acknowledged.
The acknowledgement drops by itself when the check is OK again, and a new problem alerts again. Take back starts the repetitions again.
Alerts
An email goes out when a problem is confirmed, after the attempts of the check, and one when it ends. Your provider decides who receives them, in the notifications (see Notifications).
Send notices, on the page of the hosts of the subscription, says whether the alerts are repeated while the problem lasts:
| Choice | What happens |
|---|---|
| Only when the state changes | An email when the problem starts and one when it ends |
| When the state changes, then every hour | Also a reminder every hour, while the problem lasts |
| At every failed run | An email every time the check finds the problem: with a check every 5 minutes, 12 emails an hour |
A host can have a choice of its own, on its Edit page; As set for the subscription follows the one of the subscription.