Hosting

Uptime: 10 Essential Checks for Better Website Reliability

What Is Uptime – What Is Uptime? And Why 99.9% Isn't Always Enough

Uptime describes availability over a stated period using a stated definition of success. A percentage is incomplete without knowing which service was checked, from where, how often and what counted as failure. A server that responds is not necessarily a website that lets customers complete the task they came to do.

This guide explains uptime for small business websites and WordPress teams. It includes transparent arithmetic and illustrative scenarios, not monitoring results measured on Hostreta or a claim about a particular provider’s reliability. Use the method to define your own checks, interpret a service agreement and prepare a proportionate response when something fails.

1. Define uptime around the service users need

Begin with the function being evaluated. A public article, a contact form, a login screen and a checkout depend on different paths through an application. The homepage may load while an order cannot be placed. If the business depends on checkout, a homepage-only check cannot establish that the business service is available.

Write a success condition for each important journey. A simple public page check might require an expected status and a stable piece of content. A more complex synthetic check can verify a harmless sequence using an authorized test account. Do not create real orders or send customer notifications merely to make a monitoring dashboard look complete.

Uptime should be reported at the scope actually measured. If you check one public URL from two locations, describe that scope. Do not call it proof that every feature works for every country, device and account state. Our monitoring tools guide helps turn these requirements into a suitable check configuration.

2. Uptime percentage: use an explicit denominator

For a time-based calculation, divide the available time by the total time in the reporting period and multiply by 100. Equivalently, subtract downtime from the period before dividing. Decide how overlapping incidents, maintenance and missing measurements will be handled, then apply the same rule consistently.

For example, a 30-day period contains 43,200 minutes. At a time-based target of 99.9%, the corresponding unavailable allowance is 43.2 minutes. That is arithmetic based on a declared period, not a prediction that a provider will have exactly that much downtime. A 31-day month has a different denominator.

A request-based availability calculation instead measures successful eligible requests divided by all eligible requests. It can differ substantially from time-based uptime when failures occur during a busy period. Neither calculation should be silently substituted for the other. Choose a definition that helps the team understand the user’s experience.

3. Convert percentages without hiding the period

The table below uses a 30-day month and a 365-day year, with no exclusions. It translates a target into a mathematical downtime allowance. Values are rounded for readability. It does not describe actual incidents, a guaranteed outage schedule or the terms of any specific hosting contract.

Availability target Downtime in 30 days Downtime in 365 days
99% 432 minutes 5,256 minutes
99.9% 43.2 minutes 525.6 minutes
99.95% 21.6 minutes 262.8 minutes
99.99% 4.32 minutes 52.56 minutes

The formula is total minutes multiplied by one minus the target expressed as a fraction. For 99.9%, the fraction is 0.999, leaving 0.001 of the period. Avoid mixing a year’s allowance with a monthly contractual target: a provider’s agreement may evaluate each calendar month separately.

Uptime arithmetic also says nothing about the distribution of failures. Several short interruptions and one long interruption can produce the same time-based percentage but create different recovery and communication challenges. Preserve incident history alongside the summary number.

4. Distinguish SLI, SLO and SLA

A service level indicator is the measurement, such as the proportion of successful eligible requests. A service level objective is the target chosen for that measurement. A service level agreement adds contractual terms, which can include remedies and exclusions. These concepts are related but not interchangeable.

For uptime planning, choose an objective that reflects customer needs and the team’s ability to operate the service. More nines can require additional complexity, staffing and cost. Do not adopt an ambitious number merely because another company’s marketing page displays it. Define which failures matter and how the team will act on them.

The primary Google SRE explanation of service level objectives is useful background. A small website does not need to copy a large engineering organization’s full process; it does need a shared definition of success and a reliable way to detect a meaningful failure.

5. Read the actual service agreement

Check the service named in the agreement. A commitment about a network, server platform or CDN does not automatically cover every plugin, external payment service or DNS mistake in your application. Record the measurement source, reporting period and exclusions before deciding what the advertised number means for your business.

Read the remedy and claim process. Determine whether compensation is a service credit, how it is calculated, whether it is capped and what evidence and deadline apply. Do not assume that lost revenue is covered or that a credit appears automatically. The actual agreement controls the provider’s commitment.

For illustration, Cloudflare’s Business SLA publishes service-specific conditions and exclusions. It is an example of why reading the named agreement matters, not a statement that the same terms apply to a shared hosting account or to another Cloudflare plan.

Keep contractual uptime separate from your own user-facing measurement. Both can be useful even when they differ. Explain the scope difference before accusing a provider of breaching an agreement based only on a monitor that uses another definition.

6. Build uptime checks that reflect your audience

Use an external monitoring service rather than relying only on a process running on the same server. If the server or its network fails, an internal process may not be able to send the alert. Choose monitoring locations that help identify failures relevant to your audience without claiming that a few probes represent every customer.

Set a check interval and confirmation rule that match the service’s importance. Shorter intervals can detect some failures sooner but can also produce more requests and noise. A confirmation from another location can help distinguish a local probe problem from a broader outage, while adding some detection delay.

Uptime checks should specify timeout behavior, redirects and expected content. A response with status 200 can still contain an application error or an unintended login page. Conversely, a deliberate redirect can be a valid part of the journey. Do not classify every non-200 response as an incident without considering the endpoint’s intended behavior.

7. Uptime monitoring and dependency boundaries

DNS, certificate validity, origin reachability and application behavior can fail independently. A CDN may serve a cached page while the origin is unavailable. A certificate renewal problem may affect some clients before an HTTP-only monitor notices. Use complementary checks when the dependencies matter to the service.

If a security system blocks a monitor, investigate the exact request and the provider’s documented monitoring configuration. Do not disable authentication, WAF protection or broad firewall rules simply to turn a dashboard green. Any narrow exception should be justified, reviewed and limited to the required scope.

Keep secrets out of public monitoring URLs and status pages. Use the monitoring platform’s supported credential handling when authentication is necessary, and avoid granting a monitor more access than the test requires. A reliable check should not create a new route to customer information.

8. Uptime alerts need an owner and an action

An alert that nobody receives or understands does not improve incident response. Define who is responsible during ordinary hours and outside them, which channel is appropriate and when escalation occurs. Test the delivery path with a controlled notification instead of waiting for the first production incident.

Write a short runbook for each important alert. Include the affected service, first verification steps, relevant status pages, support contacts and a recovery boundary. Avoid a generic instruction to restore a backup for every outage. A network incident and a broken application update require different actions.

Review noisy alerts without hiding genuine failures. A threshold can be too sensitive, but it can also reveal an intermittent problem that users experience. Investigate the pattern and adjust the check based on evidence. Uptime reporting should not improve merely because inconvenient failures were redefined out of the metric.

9. Confirm and contain an uptime incident

Record the first observed failure, the check result and the relevant timezone. Confirm from another appropriate path when possible and inspect provider status information. Determine whether the problem affects the whole site, one function or one region. Keep the distinction between suspected and verified causes in the incident notes.

Review recent authorized changes, domain and certificate status, application errors and relevant resource metrics. Preserve logs needed for diagnosis. Do not restart services, disable plugins or restore an old database without considering the scope and consequences. If compromise is suspected, use the appropriate security response rather than treating it as a routine performance issue.

The terminal basics guide explains how to inspect relevant information carefully, while the WordPress recovery workflow addresses suspected compromise. Neither requires assuming that every outage is an attack.

10. Recover, communicate and verify uptime

Assign one incident owner to coordinate the response and avoid conflicting changes. Choose the least disruptive supported recovery action for the evidence available. If a rollback is appropriate, confirm that it will not overwrite new orders or submissions. A previous code release and an old database snapshot have different effects.

After the site responds again, validate the affected journey and monitor for recurrence. A homepage loading once does not prove that checkout, email or account access recovered. Record the restoration time using the stated definition and note any period of degraded service separately when relevant.

Use our website incident communication guide to provide factual updates. State the verified impact and the next update time instead of promising an unconfirmed restoration deadline. Do not publish customer data or technical details that create an unnecessary security risk.

Interpret uptime reports with their limitations

Periodic uptime monitoring samples the service. A short interruption between checks can be missed, and a failure detected at one check does not establish the exact second it began. Record the interval and confirmation method. Do not report precision that the collection process cannot support.

Missing observations need explicit treatment. If the monitoring service itself failed, that gap is not proof that your site was up or down. Show the unknown period and avoid silently treating it as successful time. Likewise, a new monitor’s first week cannot establish a year’s historical reliability.

When comparing months, use complete periods and keep the endpoint, regions and rules consistent. If you change the method, annotate the report. Uptime can appear to improve because the monitor stopped checking a difficult journey, even though the customer experience did not improve.

A small-store example with no invented business results

Imagine a store whose public pages remain cached while its payment callback fails. The homepage monitor stays green, but orders do not complete correctly. The lesson is not that homepage monitoring is useless; it is that the business needs another carefully designed signal for the transactional dependency.

Use a safe test environment or a supported synthetic workflow to validate the journey without creating real customer charges. Combine monitoring with application error evidence where appropriate. A request’s status alone may not show whether a downstream action completed, so define the intended result clearly.

During an incident review, record what the check detected, what it missed and which change would improve detection. Do not invent lost sales to make the report appear urgent. If revenue impact can be measured from reliable records, state its method and uncertainty separately from the uptime percentage.

Monitoring and hosting recommendations

Promotional recommendation: KAWEDMA’s 360 Monitoring service is an option to investigate for website or server monitoring. Confirm the selected plan’s check types, locations, intervals, alert channels and retention. Monitoring can identify symptoms; it does not guarantee prevention of outages or repair the application automatically.

Affiliate disclosure: Hostreta may earn a commission if you purchase through this hosting.com managed VPS referral link. Consider it only if a managed server matches the workload and operational responsibilities. A different hosting plan is not a universal fix for DNS mistakes, application bugs or unavailable external services.

Uptime questions

Does 100% in a dashboard prove perfect reliability?

No. It means the configured measurement recorded no qualifying failure during its period. Scope, sampling and unknown gaps still matter.

Is an SLA the same as a backup?

No. An agreement describes a commitment and remedy; a backup supports recovery of data. Neither replaces monitoring or a tested incident process.

Should every website target the most nines possible?

No. Choose an objective based on customer needs, failure consequences and an operating model the team can sustain. Additional complexity can create new failure modes.

What should an uptime report include?

State the period, measured service, success rule, check method, incidents, unknown gaps and any method changes. Keep actual observations separate from contractual claims and illustrative arithmetic.

Uptime: define the measurement
Uptime: define the measurement. Original explanatory diagram.
Uptime: prepare the response
Uptime: prepare the response. Original explanatory diagram.
Uptime: verify business recovery
Uptime: verify business recovery. Original explanatory diagram.

A monthly uptime review that leads to useful work

At the end of a complete month, start with the incident list rather than the headline percentage. Group events by the affected function and the evidence available. Distinguish a confirmed application failure, a regional connectivity problem, a planned maintenance period and a gap in monitoring. These categories help the team identify repeated weaknesses without assuming every alert had the same cause.

For each incident, record detection time, acknowledgment, customer impact, recovery action and validation. A short timeline can reveal that most disruption came from delayed detection or uncertainty over ownership rather than the repair itself. Do not invent timestamps from memory when the monitoring record is incomplete; label estimates and preserve the source of each observation.

Choose one or two practical follow-up actions with an owner. Examples include testing the alert channel, documenting a rollback, adding a check for a failed form or reviewing a recurring resource constraint. A long unowned list is less useful than a small set of changes that can be completed and verified. Keep improvements separate from a promise that another incident cannot occur.

Review the business objective alongside the technical measurement. If the owner cannot tolerate losing a day’s orders, an apparently attractive hosting availability figure does not solve the recovery requirement. If the site is informational and the team has no overnight operations staff, the response model may be different. State the tradeoff rather than applying one target to every website.

Finally, record any change in monitoring configuration before the next report begins. Adding a checkout check can make reported availability look worse because the scope is now more complete. That is useful visibility, not necessarily a deterioration caused by the monitor. A transparent report explains why the number changed and what the team learned.

This review is also a good time to verify account ownership, billing contacts and domain renewal responsibility. An avoidable expiration can interrupt a technically healthy application. Keep these operational checks in the maintenance process, with an appropriate reminder and a named owner, instead of assuming the hosting provider controls every dependency.

Z
zasma171

Writes for Hostreta about the topics covered in this article.

Comments (0)

Your email address will not be published. Required fields are marked *

Ad
Affiliate links
Tools we recommend
Protect Your Website with Automated Daily Backups
$2.09
Get it
the easiest way to manage all of your social media accounts
$29
Get it
Starter Website Get online, properly
$249
Get it
google workspace Business Starter
$3.75
Get it
Nord Vpn Cybersecurity. Built for Everyday Use
$8.99
Get it
More hosting options · affiliate
AI SiteBuilder
View plans
AI Application Hosting
View plans
Shared Hosting
View plans
Reseller Hosting
View plans
Managed WordPress · Rocket.net
View plans
Unmanaged VPS Hosting
View plans
Managed VPS Hosting
View plans
Virtual Dedicated Servers
View plans

We may earn a commission on purchases made through these links, at no extra cost to you.

Ad
Ad

$ top hosting picks

affiliate
★★★★★

KAWEDMA Website Services

See optionsCheck current plans and terms
  • Explore website setup services
  • Choose a scope that fits your project
  • Confirm features with KAWEDMA
Explore KAWEDMA →
affiliate link · we may earn a commission
★★★★★

Managed WordPress Hosting

See plansCurrent terms on provider website
  • Managed hosting for WordPress
  • Powered by Rocket.net
  • Compare current plan specifications
View WordPress plans →
affiliate link · we may earn a commission
★★★★★

AI SiteBuilder

See plansCheck current offer and renewal terms
  • Explore AI-assisted website creation
  • Review included tools and features
  • Choose a plan for your project
Explore SiteBuilder →
affiliate link · we may earn a commission

Never miss a new article

Join thousands of developers and builders getting a weekly digest of the best tech articles.