Página Manutenção HTML

Maintenance pages done right: status, headers and communication

A maintenance page is what visitors see while a planned outage is in progress — a database migration, a major deploy, a payment-provider switch. The page itself is HTML, but the detail that almost everyone gets wrong is the HTTP status code. Returning 200 OK tells search engines that your maintenance copy is the new content, and Google will happily start indexing "We will be right back" as the title tag of your home page. The correct response is 503 Service Unavailable together with a Retry-After header — Google explicitly documents this as the signal to pause crawling without losing rankings.

A solid maintenance page contains a short explanation, an honest ETA, an apology, contact info and ideally a link to an external status page (Statuspage, Better Stack, or open-source Cachet) so the page stays reachable even if your main app is down. The Retry-After header accepts a number of seconds (Retry-After: 3600) or an HTTP-date — both are valid per RFC 9110.

Real-world examples and patterns

Twitter's "Fail Whale" mascot (2008-2013) turned downtime into part of the brand identity. Cloudflare's orange-cloud error page is so familiar that users immediately know the upstream is the problem. AWS publishes a Health Dashboard that is independent from the services it monitors. The cultural lesson, codified in Google's Site Reliability Engineering book, is that maintenance windows must be planned, announced and post-mortemed.

Modern deploys: do you even need this page?

Blue-green deployments, canary releases and database online-schema-change tools (gh-ost, pt-online-schema-change) make it possible to ship most changes with zero downtime. A maintenance page should be reserved for irreducible outages — a storage tier swap, a region migration, a regulatory shutdown. If you find yourself showing this page every Tuesday at 02:00, the maintenance is not the problem; the deploy pipeline is.

Communication and SLAs

For B2B SaaS the maintenance window has contractual consequences: most SLAs deduct planned-maintenance hours from the 99.9% target, but only if the customer was notified in advance (often 72 hours). Combine the on-site banner with email to admins, a Slack/Teams notification, a tweet from @YourCompanyStatus and an incident channel internally. Post-mortem within 48 hours of resolution — even for planned downtime — is the modern standard.

FAQ

Why is the 503 status code so important? Because 200 on a maintenance page lets Google re-index your site with the placeholder copy and tanks rankings. 503 + Retry-After is interpreted as "temporarily down, come back later" and preserves the index.

How long can a maintenance window last? Under two hours is ideal for consumer products; under four hours for B2B. Beyond that, fragment the work into smaller windows or invest in zero-downtime tooling — long single-window maintenance is a smell, not a strategy.

Should I offer an alternative to users? Yes when possible. A link to a read-only mirror, a status page, a Twitter feed or even a competitor courtesy link signals respect for the user's time and reduces support load.

WordPress, Shopify and Webflow have built-in maintenance modes — do I need a custom page? Built-ins are fine for small sites. For anything where SEO or branding matters, ship a custom HTML page served by your CDN (Cloudflare Workers, Vercel) so it stays up even if the origin is offline.