Skip to content
HonkWatch
PricingLocationsHonk RadarHelpAPI Log inGet started
All help topics

Getting started

  • Getting started
  • How checks work

Monitors

  • Which monitor to use
  • Website monitor
  • API monitor
  • Keyword monitor
  • Ping monitor
  • Port monitors (TCP and UDP)
  • DNS monitor
  • SSL certificate monitor
  • Domain expiry monitor
  • Third-party services
  • Page load monitor
  • Heartbeat monitor (cron jobs)
  • Maintenance windows
  • Incidents

Alerts

  • Setting up alerts
  • Email alerts
  • Slack
  • Microsoft Teams
  • Discord
  • Telegram
  • PagerDuty
  • ntfy
  • Webhooks
  • Zapier

Phone and desktop app

  • The phone and desktop app
  • App alerts not arriving

Status pages

  • Status pages
  • Status page on your own domain

Team and account

  • Team members
  • Plans and limits
  • API keys and the API
  • Your account

Troubleshooting

  • A monitor says down but my site works
  • I did not get an alert
  • Allowing our checks through a firewall
  • Questions and answers

Help / Getting started

How checks work

This page explains where our checks come from, how often they run, and the rule we use before we call a monitor down and send an alert. Knowing the rule helps you tune a monitor so it is neither noisy nor slow to warn you.

Where checks come from

Checks run from servers in cities around the world. The Locations page lists them. Each check is made over the public internet, the same way a visitor would reach you, which is why only public addresses can be monitored.

Some cities have more than one checking server in the same data centre. For deciding whether something is down, servers in one data centre count as one city, because they share the same network.

How often

You choose the interval with Check every when you add or edit a monitor. The choices are 30 seconds, 1, 2, 5, 10, 15 and 30 minutes, 1 hour and 6 hours. The form only lists the intervals your plan allows. Plans and limits gives the fastest interval for each plan.

A few monitor types have their own floor:

  • Page load monitors run at most every 5 minutes.
  • Domain expiry monitors run every 6 hours, because registries limit how often they can be asked.

Within an interval, checks from different cities are spread out over up to two minutes rather than all arriving at the same second. This keeps small servers and firewalls from seeing a burst of requests.

Which cities check a monitor

Open More settings on the monitor form to see the Check from list.

  • No city ticked (the default): all cities take part, taking turns. Each round only a few of them check, and over a number of rounds every city has a go.
  • Cities ticked: only those cities check. If you tick three or fewer, all of them check every round. The label above the list shows how many cities your plan lets you tick.

When we call it down

One failed check is not enough. Two settings under More settings decide when a monitor counts as down:

SettingDefaultRange
Down when this many locations agree21 to 10
Each of them after this many failed checks in a row21 to 3

With the defaults, a monitor goes down when two different cities have each failed twice in a row. The form spells the rule out in one sentence under the city list, in the form "Alert when any 2 of ... fail twice in a row." The sentence changes as you change the settings.

Here is what happens when a city sees a failure:

  1. That city checks again about 28 seconds later instead of waiting a full interval.
  2. While the monitor is still up, other cities are asked to check at once, starting with one in the same part of the world and one elsewhere. If the failures go on, more cities in the same part of the world are asked. This is how an outage that only one region can see still gets confirmed.
  3. When enough cities have failed enough times in a row, the monitor goes down, an incident opens and the down alert goes out. The alert says what failed and which cities saw it.

If you ask for more agreeing locations than there are cities checking the monitor, we use the number of cities available instead.

A city whose last answer is old does not vote. An answer counts for three intervals (at least 90 seconds). After that the city shows as Unknown on the monitor page, because a quiet checking server tells us nothing about your site.

When it comes back up

A down monitor stays down until at least one city gets a good answer and fewer cities than the number you set are still failing. Then it comes back up. The incident closes and a "back online" alert goes out, saying how long it was down and what the cause was.

You can turn the recovery message off for one monitor: untick Send a "back online" message when it recovers under More settings. Each alert channel also gets back up alerts only if Back up was ticked when it was added on the Alerts page.

Timeouts

Each check waits up to the monitor's Timeout (seconds) for an answer. The default is 10 seconds and you can set 1 to 30. A check that gets no answer in time counts as a failure.

Other alerts a monitor can send

  • Repeat alerts. Set Repeat the down alert every (minutes, 0 = once) to keep reminding you during a long outage. The shortest repeat is every 5 minutes. Repeats stop when someone presses Acknowledge on the incident, on the monitor's page.
  • Slow. Set Alert when slower than (ms, optional). By default we compare the average time of the cities with a recent good answer against your limit. Tick Alert when any single location is slower to compare the slowest city instead. You get one slow alert, then another only after it has been back under the limit.
  • Certificate expiring. Website and keyword monitors on https:// addresses read the certificate as they check. You get a warning when it has fewer days left than Warn before certificate expiry (days), 14 by default. An SSL certificate monitor is stricter: it counts as down when the certificate is invalid or has fewer days left than that setting.

Monitors that work differently

  • Domain expiry and Third-party service monitors are checked from one place each round, because the answer is the same from everywhere. The location settings above do not apply to them.
  • Heartbeat monitors are not checked by us at all. Your job calls us; if no call arrives within the expected time plus the grace period, the monitor goes down. See Heartbeat monitor.

Maintenance and paused monitors

A paused monitor is not checked and sends nothing. During a maintenance window, a new outage does not open an incident or send a down alert. If the monitor is still down when the window ends, the incident opens and the down alert goes out then. See Maintenance windows.

If checks do not behave as you expect

  • Down alerts for a site that works: see A monitor says down but my site works. Raising the number of agreeing locations, or the failed checks in a row, makes alerts slower but quieter.
  • Your firewall may block our checking servers. Allowing our checks through a firewall lists what to allow.
  • An alert did not reach you: see I did not get an alert.
← Getting started Which monitor to use →

Still stuck? Email [email protected] and say which guide you were following.