Langues

Guide Website Hosting Migration SEO

Déplacer a website to a nouveau host, CDN, or DNS provider sans modification URLs: preparation, cutover, validation, monitoring, and rollback.

Première publication : 18 juil. 2026 · Dernière mise à jour : 3 août 2026 · Advanced
Langues
AIRédigé parAI
1 indice probant sur cette page

A hosting migration changements the infrastructure behind a site pendant que keeping its public URLs stable. Preserve the même content and SEO signals, lower DNS TTL ahead of the cutover, prove the nouveau origin and CDN peut serve utilisateurs and verified robots d’exploration, run old and nouveau infrastructure in parallel, comparer réponses and rendered pages, monitor les deux sets of logs, and retire the old host seulement après its trafic reaches zero. Redirection maps and Modifier of Adresse ne sont pas partie of a vrai same-URL hosting déplacer.

TL;DR — Treat a same-URL hosting, CDN, or DNS migration as une réponse-parity and traffic-routing project. Inventory every hostname and dependency, lower DNS TTL avant launch, configurer the nouveau origin and edge, validate certificates and security contrôle, load-test realistic robot d’exploration and utilisateur demand, and comparer raw plus rendered réponses. Dual-run old and nouveau infrastructure via DNS propagation. Monitor les deux log streams, DNS réponses, errors, latency, cache behavior, explorer activity, and Search Console. Roll back by restoring the previous routing seulement quand a pre-agreed infrastructure échec occurs.

Decide si ce is really a same-URL migration

A same-URL hosting migration changements infrastructure sans modification the exact public URL string. The scheme, hostname, port, chemin, requête handling, and trailing slash behavior remain stable.

Classify the project avant planning it:

ModifierSame-URL hosting déplacer?Additional migration fonctionner
Nouveau origin IP, même URLsYesRéponse parity, DNS, capacity, logs
Nouveau CDN, même URLsYesEdge rules, cache, TLS, firewall, origin routing
Nouveau authoritative DNS providerUsuallyZone parity, delegation, DNSSEC, mail and service records
www.example.com to example.comAucunURL mapping and redirection permanentes
HTTP to HTTPSAucunProtocol migration and per-URL redirections
Chemin or CMS-generated URL changementsAucunURL migration plus platform QA

Ne faites pas let a project manager étiquette une URL modifier as “just hosting.” The deployment plan doit inclure every migration type que en réalité ships.

Construire the infrastructure inventory

Infrastructure inventory empêche the quiet dependencies from becoming launch-day surprises. Record:

  • tout public hostnames, notamment assets, images, APIs, international hosts, and legacy aliases;
  • A, AAAA, CNAME, NS, SOA, CAA, MX, TXT, and relevant SRV records;
  • certificate issuers, validation méthodes, Subject Alternative Noms, and expiry;
  • origin addresses, ports, health checks, charger balancers, and failover behavior;
  • CDN cache keys, cache rules, redirections, transforms, workers, and purge méthodes;
  • WAF, bot, rate-limit, geo, authentication, and IP autoriser/deny rules;
  • réponse headers, compression, cookie behavior, and security headers;
  • log destinations, retention, sampling, fields, and temps zones;
  • Search Console and analytics verification méthodes;
  • third-party callbacks, webhooks, payment flows, feeds, and allowlisted IPs.

DNS examiner doit inclure non-web records. Breaking MX, SPF, DKIM, DMARC, or service records may pas directement modifier rankings, but it peut break the business vous were trying to protéger.

Establish une réponse-parity baseline

Réponse parity signifie comparing the old and nouveau systems pour the même requested URL, pas merely checking que les deux retourner 200.

Capture a representative définir à travers templates and behaviors:

  • status and chaîne de redirections;
  • final URL and protocol negotiation;
  • title, canonical, robots directives, hreflang, and données structurées;
  • raw HTML and browser-rendered principal content;
  • Content-Type, Cache-Control, Vary, compression, and security headers;
  • images, fonts, JavaScript, CSS, PDFs, and media assets;
  • cookies and logged-in or personalized variants;
  • mobile and desktop behavior;
  • latency, temps to premier byte, and error rate.

Utiliser the Staging vs. Production SEO Diff pour paired page checks. A complet robot d’exploration and scripted requête suite devrait cover the plus grand inventory.

Prepare the nouveau origin

Origin preparation starts with content and configuration parity. Copy current content, templates, media, robots rules, redirections, error handling, and verification fichiers. Freeze or synchronize writes so the nouveau database ne fait pas launch stale.

Tester the origin directement via a controlled hostname, a local hosts-file override, or provider-specific preview mechanism. The tester doit preserve the production Host header parce que virtual hosts, application routing, certificates, canonicals, and absolute liens souvent depend on it.

The nouveau origin doit aussi handle the post-cutover charger. Warm the application and database, confirmer connection pools and autoscaling, and load-test uncached demand. CDN cache misses peut concentrate trafic at the origin immédiatement après launch.

Configurer the CDN as a separate system

CDN migration changements plus que geography. Comparer the old and nouveau edge behavior explicitly:

  • cache clé composition, notamment requête strings, cookies, headers, and device variants;
  • cacheable code d’états and fichier types;
  • navigateur TTL, edge TTL, stale serving, revalidation, and origin shielding;
  • redirections, rewrites, header transforms, and edge functions;
  • cache bypass rules pour accounts, carts, search, and personalized pages;
  • compression and image optimization;
  • purge scope and propagation;
  • WAF, bot management, rate limiting, and origin protection.

Cloudflare’s current documentation, Par exemple, notes que its par défaut mise en cache may respect origin Cache-Control headers but peut be overridden by edge rules. It aussi provides targeted or complet purges to force fresh origin récupère. The exact behavior is vendor-specific, so export and comparer configuration plutôt que assuming equivalent étiquettes mean equivalent results. Voir Cloudflare’s cache documentation.

Treat cache parity as content parity

Cache configuration peut serve the incorrect page correctement and quickly. Tester anonymous, authenticated, localized, mobile, and query-string variants. A cache clé que omits a meaningful cookie or header peut leak personalized content. A cache clé que inclut every tracking parameter peut fragment the cache and overload the origin.

Purge or pre-warm critical assets and pages according to the launch plan. Ne faites pas blindly purge everything during peak trafic unless the origin has been testé pour le résultating miss storm.

Validate TLS from utilisateur to edge and edge to origin

TLS validation has two legs quand a CDN terminates HTTPS: navigateur to CDN and CDN to origin. Confirmer hostname coverage, complet certificate chains, modern protocol prise en charge, renewal, and strict origin validation.

Origin-only certificates may pas be publicly trusted. Cloudflare warns que its Origin CA certificates peut produce navigateur trust errors si proxying is disabled or paused. Que matters during rollback: a DNS-only fallback to an origin en utilisant an edge-only trust model may échouer pour utilisateurs. Voir Cloudflare Origin CA guidance.

Tester every public hostname, notamment wildcard assumptions and rarely utilisé asset or regional hosts. A valid apex certificate ne fait pas prove every subdomain is covered.

Lower DNS TTL avant the déplacer

TTL planning starts avant cutover. Google recommends lowering the relevant TTL to a conservative low valeur, tel as a few hours, au moins un week avant the déplacer. A DNS provider may impose différent minimums; proxied records may aussi have fixed valeurs.

Cloudflare’s TTL documentation explique the basic tradeoff: plus long valeurs augmenter cache reuse, pendant que shorter valeurs autoriser record changements to prendre effect sooner. Record the original TTL and schedule its restoration seulement après the nouveau infrastructure is stable.

DNS changements peut be non-atomic à travers distributed systems. Modifier as little as possible during the cutover, vérifier réponses from several public resolvers, and garder the old destination disponible pendant que mis en cache réponses remain valid.

Vérifier robot d’exploration accès and security contrôle

Security parity n’est pas rule-count parity. A WAF copied from un autre provider peut challenge or block robots d’exploration, strip requête parameters, rewrite réponses, or rate-limit high-volume exploration differently.

Google’s hosting guide dit to garantir firewalls and denial-of-service protection do pas block Googlebot from DNS or hosting servers. Vérifier Googlebot en utilisant Google’s documented verification méthodes, pas a user-agent string alone.

Tester les deux ordinary robot d’exploration behavior and legitimate bursts. Éviter broad allowlisting que disables protection pour spoofed utilisateur agents. Preserve security logs so blocked requêtes peut be distinguished from origin échecs.

Plan the dual run

Dual running signifie les deux old and nouveau infrastructure peut serve correct production réponses during propagation. The old environment doit garder receiving content or données changements que affecter le site. Sinon utilisateurs routed by mis en cache DNS réponses may voir stale inventories, broken sessions, or outdated pages.

Pick a synchronization strategy:

  • un lire/écrire database shared by les deux stacks;
  • replicated données with an understood lag and conflict policy;
  • a controlled content freeze during cutover;
  • one-way event replication pour orders, formulaires, or utilisateur writes.

Session state, uploads, cache invalidations, and background jobs besoin the même decision. “Both servers are on” n’est pas a dual-run plan si leur state diverges.

The old environment is a rollback path only while it remains valid and synchronized. Retirement begins when logs prove the old path is no longer used. Source : Website Hosting Migration SEO

Prepare builds the new origin and edge path. Validate tests controlled routing, parity, certificates, and capacity. Dual run keeps old and new environments correct and synchronized. Cut over changes only the planned DNS or edge route. Drain observes old-host requests in separate logs while the old environment remains available. Retire occurs only when old-host traffic reaches zero and dependencies have moved. A rollback lane remains available before retirement when a pre-agreed infrastructure failure occurs and the old state is still valid.

© Patrick Stox LLC · CC BY 4.0 ·

Execute the cutover

Hosting cutover devrait be deliberately boring:

  1. Arrêter unrelated deployments and confirmer the modifier window.
  2. Run the final parity, certificate, capacity, and backup checks.
  3. Supprimer temporary explorer or accès blocks from the nouveau production chemin.
  4. Modifier seulement the planned DNS or CDN routing records.
  5. Confirmer attendu réponses from multiple resolvers.
  6. Requête protected pages via the public route as a utilisateur and robot d’exploration.
  7. Confirmer logs are arriving from edge, nouveau origin, and old origin.
  8. Watch errors, latency, cache misses, origin charger, and conversions.

Ne faites pas utiliser Google’s Modifier of Adresse outil pour a host-only déplacer. Aucun public URL has modifié, so là is aucun adresse modifier to report.

Monitor the evidence que proves the déplacer

Infrastructure monitoring devrait separate old and nouveau trafic. Utiliser a deployment marker and comparer the même time-of-week baseline où seasonality matters.

Watch:

  • DNS réponses and resolver propagation;
  • old-host and new-host requêtes by utilisateur and verified robot d’exploration;
  • edge and origin status-code distribution;
  • TLS, connection, timeout, and application errors;
  • latency percentiles and uncached origin réponse temps;
  • cache hit ratio and origin requête volume;
  • Googlebot requêtes, Statistiques d’exploration, Page Indexation, and representative Inspection d’URL;
  • synthetic checks à travers regions and networks;
  • analytics, conversions, and critical business transactions.

Google dit a temporary Googlebot crawl-rate drop immédiatement après a hosting modifier peut be normal, followed by an augmenter over the suivant few days. Anchor quelconque decision to accessibility and error evidence, pas que attendu pattern alone.

Define rollback avant launch

Rollback renvoie routing to a known-good infrastructure state. It n’est pas a vague promise to “switch DNS back.” Document:

  • the exact records, routes, and configurations to restore;
  • who peut authorize and execute the reversal;
  • how modifié content, sessions, formulaires, orders, and uploads va reconcile;
  • si the old certificates and dependencies remain valid;
  • cache purge steps on les deux routes;
  • the échec thresholds que trigger rollback;
  • the maximum safe decision temps.

Rollback triggers devrait be observable: sustained availability échecs, material conversion breakage, widespread incorrect content, certificate échecs, robot d’exploration blocks, or capacity collapse que ne peut pas be corrected dans the window. A temporary explorer rate fluctuation by itself n’est pas a rollback trigger.

Retire the old infrastructure from logs, pas a calendar

Old-host retirement se produit après logs montrer que utilisateurs and robots d’exploration ne … plus reach it and tout dependent services have déplacé. Google recommends shutting bas the old host après its trafic reaches zero.

Retain configuration exports, logs, and rollback artifacts according to business requirements. Restore DNS TTL to the intended steady-state valeur après stability is proven. Supprimer temporary firewall exceptions and duplicate scheduled jobs so the migration ne fait pas leave a permanent maintenance mess.

Add an expert note

Pin an expert quote

New person? Create their unclaimed profile at /admin/experts/ → Pin a quote first.