Get an alert when your website goes down: a practical setup for small businesses and agencies
A visitor who finds your site down does not tell you. They leave. A monitor checks the site for you and tells you first. This guide covers what to check, how often, how to keep false alarms out and who should get the alert.
- For one site or a portfolio of client sites
Monitor what earns money, not only the homepage
A homepage that loads says little about the form that takes orders. Start with this list and remove what you do not have:
- Homepage. An HTTP check that expects status 200 and a text that only appears when the page works, such as your company name. A page can answer 200 and still show an error.
- The page that makes money. Checkout, booking form, quote form or customer login. Check it on its own, with a text that must be there, such as “Add to cart”, and a text that must not, such as “Database error”.
- API. The endpoint your app or partners call, with the method, headers and status code it should return.
- SSL certificate. An expired certificate blocks every visitor. Watch the expiry date, the chain and the host name; see how to monitor SSL certificate expiry.
- DNS. An alert when your records change. A wrong edit at the registrar can send visitors and email elsewhere while your server runs fine.
- Scheduled jobs. Backups, imports and mail queues fail without a sound. A heartbeat expects the job to call a URL on each run and alerts when it does not.
An agency repeats this per client and groups the monitors per client, so the right people see the right sites.
How often to check
The interval is the longest a failure can go unnoticed before the first failed check. For a shop, a booking system or an API, check every minute, or every 30 seconds where every minute counts. For a brochure site, every 5 minutes is enough. A shorter interval only helps if someone can act on the alert that fast.
Keep false alarms out
False alarms are the main reason people mute alerts, and a muted alert is no alert. Three settings keep them out:
- Confirm from a second location. One probe can lose its route to your server while your visitors see nothing wrong. Confirm a failure from a second region before you alert.
- Require more than one failure. Alert after two failed checks in a row, not after one slow response.
- Plan for maintenance. Put planned work in a maintenance window, so an update at night does not page anyone.
Confirmation costs a little time. That is a good trade: an alert that arrives two minutes later and is always real gets read.
Who to alert, and how
Alert a person who can act, on a channel they see within minutes.
- Email for every alert, as a record. It is too slow to wake anyone.
- Slack, Microsoft Teams or Telegram for the team during working hours. Telegram also reaches a phone.
- SMS for critical sites outside working hours. With UptimeVanguard, SMS goes through your own Vonage, Twilio or CM.com account, and your provider bills the messages.
- Recovery messages on the same channels, so nobody keeps fixing a problem that is already solved.
Agencies: route each client's monitors to that client's channel with a routing rule, instead of sending every alert to everyone. Send a test message once to prove each channel works.
Tell your customers with a status page
During an outage, customers want to know if the problem is on their side. A public status page answers that question without a support ticket. Visitors can subscribe by email and get a message when a component goes down or recovers, and when you post an update. UptimeVanguard status pages run on our address and show “Powered by UptimeVanguard”.
What it costs with UptimeVanguard
The Free plan covers 3 sites, checked every 5 minutes from one region you pick, with every check type, every alert channel and 1 status page. A second region confirms an outage before you are alerted, as on the paid plan.
The paid plan costs €10.00 a month for 5 sites, with every check type, ten regions, second-region confirmation, all alert channels and up to 5 status pages. Each extra site costs €1.50 up to 25 sites, then less. For an agency, 12 sites cost €20.50 a month, excl. VAT. A site is one monitor: its HTTP, certificate, DNS and ping checks count once. A checkout page on its own monitor is a second site. The pricing page lists every size.
You can start for free; the first check runs in about two minutes.