Uptime Monitoring: Knowing When Your Site Is Down

Why downtime matters beyond lost visits

While a site is down, enquiries and sales stop. If it stays down, search visibility can suffer too. Google's page on how HTTP status codes affect its crawlers says server errors (5xx) and 429 responses make Google crawl more slowly. Pages already in the index are kept for a while but eventually dropped, and URLs that keep returning a server error are removed. For network and DNS errors, where Google cannot reach the server at all, Google says already indexed URLs that are unreachable will be removed within days.

A short outage is unlikely to matter much, but one that runs unnoticed over a weekend can. Monitoring makes the gap between "it broke" and "someone is fixing it" as short as possible.

What a monitor checks

Google's site reliability engineering book, in its chapter on monitoring distributed systems, calls this black-box monitoring: testing externally visible behaviour as a user would see it. An uptime service requests your pages on a schedule and records what comes back. The common check types:

• HTTP status. The page should return 200. A 5xx code, a timeout or a refused connection counts as down.
• Keyword or content check. Some failures still return 200: Google notes it sees many sites that return 200 for server errors, downtime and placeholder pages. Checking that a word or phrase which should be on the page is actually there catches a blank page or an error message served as a "success".
• Response time. A site that takes many seconds to answer is effectively down for many visitors. Alert on slowness as well as failure.
• Certificate expiry. An expired HTTPS certificate makes browsers show a warning instead of your site. Let's Encrypt has ended its expiry email service, so a monitor that warns ahead of the expiry date is worth having. See our SSL certificates guide for how renewal works.
• DNS and domain. If the domain's DNS stops answering or the domain lapses, the whole site and often the email go with it. A DNS check, plus auto-renew on the domain, covers this; our guide to domain names explains renewals.

If your service can check from more than one location, use that and require more than one failed check before alerting, so a single network blip does not wake anyone up.

Which pages to watch

The home page alone is not enough. It may be served from a cache while the parts that earn money are broken. Watch:

• the home page;
• a page that is built fresh from the database or content system, not only cached pages;
• the contact or booking page, and the checkout if you sell online;
• any API or app backend that a mobile app or other system depends on.

Monitors do not submit forms or complete purchases. Keep testing those by hand after every change, as set out in our website maintenance checklist.

Alerts people will act on

An alert is only useful if the right person sees it and does something. The SRE chapter is direct about this: every page should be actionable, and people can only react with urgency a few times a day before fatigue sets in. It also separates symptoms, what is broken, from causes, why. For a business website, the symptom that matters is that visitors cannot use the site, so alert on that first.

• Decide who is told. Name a first contact and a backup, and agree what each should do.
• Use a channel that does not depend on the site. If your email runs on the same server as the website, an outage can take the alert email down too. Use a text message, an app notification or an address with a different provider.
• Send recovery messages too, so people know when to stop investigating.
• Tune noisy checks. An alert that often fires for nothing teaches people to ignore it. Raise the threshold or fix the cause rather than putting up with it.

Status pages

A status page tells customers whether something is wrong and when it is expected to be fixed, which cuts the number of "is it just me?" messages. It is most useful for web apps, portals and online services that customers log in to. Host it separately from the main site so it stays up when the site does not, keep updates short and factual, and post when the problem is resolved. A brochure site usually does not need one; a short post on your usual social channels does the same job.

What to do in an outage

1. Confirm it. Open the site on a phone using mobile data, not the office network, to rule out a local problem.
2. Check the host's status page and any notices from your hosting provider. If the problem is on their side, open a ticket and wait for their update.
3. Look at what changed. An update, a plugin, a DNS change or an expired certificate just before the outage is the most likely cause. A maintenance log makes this quick.
4. Roll back or restore if a change caused it. Our website backups guide covers how to restore safely.
5. Afterwards, write down what happened, how long it lasted and what will stop it happening again.

If the outage looks like a hack, for example with defaced pages or redirects you did not create, see our website security basics guide.

Planned maintenance without hurting search

When you take a site offline on purpose, tell search engines it is temporary. Google's guidance on planned downtime says to return a 503 (Service Unavailable) status, not a 404 or an error page with a 200 status, and suggests an optional Retry-After header with the expected length of the downtime, which Googlebot may use to decide when to come back.

Google's page on pausing an online business adds limits: use a 503 page only for an urgent shutdown of one to two days, do not return 503 for the robots.txt file because that blocks all crawling, and use a normal indexable placeholder page if the site must be off for longer. Pause your uptime alerts for the maintenance window, then switch them back on.

Frequently asked questions

What is uptime monitoring?

A service that requests your website from outside on a schedule, as a visitor would, and alerts someone when it fails, returns an error, answers too slowly or is missing expected content.

Is checking the home page enough?

No. The home page may be cached while the database, contact page or checkout is broken. Also watch a page built fresh from the database, the pages that bring in enquiries or sales, and any API an app depends on.

Does downtime hurt my Google rankings?

A short outage is unlikely to matter much. Google says server errors make it crawl more slowly, and URLs that keep returning errors, or that it cannot reach at all, are eventually removed from its index.

What should my site show during planned maintenance?

A short page with a 503 (Service Unavailable) status, ideally with a Retry-After header. Google recommends this only for one to two days, and says not to return 503 for robots.txt.

Want someone else to watch your site? Our website maintenance service includes uptime and server health checks, updates and backups. See all our services, or email info@creahive.com.