Skip to content
Technology & operations

Switching Web Host: Move Site and Email Without Downtime

Change hosting provider without downtime for site and mailboxes: backups, auth code, TTL, parallel operation and a migration schedule from T-14 to T+7.

13 min read WebhostingDomainumzugDNSE-Mail-ZustellbarkeitWebsite-Betreuung

A provider change rarely starts with a technical wish. Usually it is a price that no longer fits, a contact who stopped answering, or a server that has become too slow. The move itself is then treated as a formality: copy the files, import the database, switch the name servers. In practice a change rarely fails on the technology, it fails on the order of steps. Cancel before the domain has moved and you lose the convenient route to the auth code. Switch the name servers without lowering the lifetime of the DNS records first and you wait hours for a resolution that should have changed long ago. Forget the mailboxes and you notice only when order confirmations and enquiries stop arriving. More than 18 million (DENIC) .de domains are registered in Germany, and every one of them sits on a contract that ends at some point. This article describes the provider change as a project: what has to be secured before you cancel, how domain, DNS, website and email move separately from each other, why certificates, redirects and cron jobs are the most common break points, which rights to source code and content belong in the contract, and what a schedule from T-14 to T+7 looks like that keeps the actual cutover down to a few minutes.

Key takeaways

  • A provider change consists of four separate parts: domain, website, mailboxes and DNS zone do not move together, they move one after another.
  • Before you cancel, the database, media files, mailbox contents, credentials and the domain auth code have to be secured and checked for restorability.
  • A lowered lifetime on the DNS records, genuine parallel operation and a local test resolution keep the cutover window down to a few minutes.
  • The most common break points are not the contents but certificates, redirects, form delivery and cron jobs that were anchored in the old environment.
  • Rights of use for source code, images and text belong in the contract; without an explicit grant the scope stays limited to the agreed purpose.
  • After a change of sending address, mail reputation needs a run-up: raise volume gently, move authentication along, watch delivery rates.

The change is a project, not a button

Anyone switching provider is moving not one thing but four: the domain with its registration contract, the website with files and database, the mailboxes with their contents, and the DNS zone that ties everything together. These four parts belong together, yet they have to be handled separately in technical and contractual terms. Experience shows that a move rarely fails because a database dump goes wrong; it fails because somebody tries to move all four at once. What makes it harder is that responsibilities are often scattered: the domain is registered in the name of a former agency, the mailbox runs on a different contract, and the server credentials are known only to a colleague who left the company long ago. The first step of a change is therefore not a technical action but an inventory.

That inventory is larger today than it was ten years ago, because more parts are outsourced. According to the survey on the use of information technology in enterprises, around 54 percent (Federal Statistical Office) of companies with ten or more employees used paid cloud services in 2025, and among small businesses with ten to 49 employees the figure was 51 percent (Federal Statistical Office). Each of those services brings its own credentials, its own notice periods and its own export routes. At the same time the infrastructure in the country keeps growing: Germany operates around 2,000 data centres (Bitkom) with more than 100 kilowatts of connected load, and their total capacity rose to 2,980 megawatts (Bitkom) in 2025, up 9 percent (Bitkom) year on year. If you want to weigh up the merits of a domestic location, the legal and technical arguments are in our article on web hosting in Germany; the practical side is described on our page about hosting in a German data centre.

Four parts that move separately

The domain is a contract with a registry and moves via an auth code. The website is a directory tree plus a database and moves via copy and import. The mailboxes are a service of their own and move via a synchronisation of messages. The DNS zone is the control desk and decides which of the two setups is currently served. Keep these four apart and you can move them one after another while keeping a way back at every stage.

What has to be secured before you cancel

As soon as a contract is cancelled, a clock starts running at the end of which data is deleted. That is exactly why the backup belongs at the beginning and not at the end. It is complete only when three things exist: a plain-text dump of the database, an image of the entire directory tree including media and configuration files, and a list of all credentials. A dump alone is not enough. The German Federal Office for Information Security explicitly recommends in its data backup module that restoration be tested at regular intervals (BSI), and that applies doubly before a move. A backup that cannot be imported on the new server is not a backup, it is a file. So check before you cancel whether the dump can be turned into a working setup, and write down which steps that required.

Mailbox contents are part of the backup too. When a mail account is switched off, the stored messages are usually lost, and with them part of the business correspondence that commercial and tax law requires you to retain. If you have engaged a processor, you can additionally rely on Article 28(3) (GDPR): the contract must provide that personal data is returned or deleted after the end of the service, at the controller's choice. For data based on consent or on a contract, the right to data portability under Article 20 (GDPR) applies as well, which requires a structured, commonly used and machine-readable format. Our checklist for a privacy-compliant website summarises the records and registers you should be keeping anyway.

Database and files

A plain-text dump of the database, plus the complete directory tree with media, configurations and custom adjustments. Only a test import in the new setup proves that the state is complete and runnable.

Mailboxes and distribution lists

Messages, folder structure, address books, forwards, shared mailboxes and out-of-office rules. Without this list you will be missing exactly those addresses that are used rarely and still matter.

Credentials and zone file

Server access, database account, registry, auth code, certificate management and a printout of the complete DNS zone with all records, including the seemingly unused verification entries.

  • Create a database dump and import it into the new setup as a trial
  • Copy the directory tree in full, including hidden configuration files
  • Take over all mailboxes with folder structure, address books and rules
  • Document the complete DNS zone, including records for verification procedures
  • Request the domain auth code and note how long it stays valid
  • Collect credentials for interfaces, keys and connected services
  • Draw up a list of all cron jobs and scheduled tasks with their start times

Domain, auth code and transfer deadlines

The domain is the only part a company can genuinely lose. It sits on a contract with a registry, and changing the provider that looks after it follows a defined procedure. For .de domains it is called a provider change and relies on a stored password, the AuthInfo. Since 2 February 2010 (DENIC) this is the only procedure applied: the outgoing provider deposits the code with the registry, where it stays valid for 30 days (DENIC). If the submitted code matches the deposited one, responsibility changes within minutes, without touching name servers or content. With a stock of more than 18 million (DENIC) .de domains this is routine, yet it regularly fails on outdated contact details: if the register lists an address that no longer exists, the code does not arrive.

For generic endings such as .com or .net, the ICANN transfer policy applies. It obliges the outgoing provider to respond to a transfer request within five days (ICANN) as a rule, and it locks a domain for 60 days (ICANN) against a further provider change after a transfer or a change of registrant. That lock is the reason a move cannot be squeezed in at short notice. On top of it come transfer locks that many providers set by default and that have to be lifted before the move. So plan the domain part as the first task, not the last. Reverse the order and you get the same surprises as in a poorly prepared rebuild, which we described in our article on the website relaunch without ranking loss.

Move first, cancel afterwards

A cancellation ends more than the hosting; it often ends the old provider's willingness to hand over the auth code or keep the mailboxes running. So cancel only once domain, data and mailboxes sit in the new environment and have been checked. Keep an eye on the notice period so the contract does not renew unintentionally, and still leave an overlap of a few weeks. Those weeks cost little and are the cheapest way back that a move offers.

Lower the record lifetime, run in parallel, cut over

The actual cutover happens in DNS. Every record carries a value for how long it stays valid, the time to live, defined in the standard as a figure in seconds (IETF RFC 1035). Until that period has elapsed, caching name servers keep handing out the old answer, regardless of what the zone already contains. Common default values are 3,600 or 86,400 seconds, that is one hour or one day. If you plan a cutover without downtime, you lower the lifetime of the affected records to around 300 seconds about 48 hours in advance and then wait out the old value once in full. After that the network reacts to a change within five minutes instead of within a day, and the way back to the old environment is just as fast.

A second point is often overlooked: the caching of negative answers. When a name server queries a record that does not exist yet, it remembers that absence as well. The standard derives the duration from the last field of the SOA record and recommends capping it at three hours (IETF RFC 2308). So if you create a new name after having accidentally queried it, you wait longer than necessary. Create new records before anyone looks for them, and accept the new setup through a local test resolution in which only your own machine talks to the new server. That way you see the complete state under the real address while everyone else keeps getting the old site unchanged. The fact that a setup has to be watched after the move applies just as much as in the day-to-day care of a maintained website.

  • Lower the lifetime of all affected records to around 300 seconds 48 hours in advance
  • Build the new setup in full and accept it through a local test resolution
  • Pause write access briefly so orders and forms do not land on two different states
  • Synchronise the database and changed files one last time
  • Switch the records in the zone and check resolution from several networks
  • Raise the record lifetime again once stable operation has been demonstrated

Why the cutover itself takes only minutes

Downtime during a move arises almost entirely in the gap between two states: the old environment is switched off and the new one is not reachable yet. That gap can be avoided by letting both states run in parallel for a while and letting DNS decide which one is served. The lowered record lifetime makes that decision reversible. What remains in practice is a short window of a few minutes in which write access pauses so no order ends up on the wrong state.

The most common break points after the cutover

Certificates are the break point that shows up fastest, because the browser puts up a warning page. In the old environment the certificate was renewed automatically; in the new one that automation has to be set up first, and it only works once the domain already points to the new server. The deadlines are getting shorter: the CA/Browser Forum has resolved to cap the maximum lifetime of publicly trusted TLS certificates at 200 days (CA/Browser Forum) from 15 March 2026, at 100 days (CA/Browser Forum) from March 2027 and at 47 days (CA/Browser Forum) from March 2029. A working automatic renewal is therefore no longer a convenience but a prerequisite for operation. Check in addition whether strict transport security is active, because then the browser will not forgive even a short gap. What else belongs to technical hardening is set out in our website security basics.

The second break point is redirects. They often live in a server configuration that does not come along when the directory tree is copied, because in the old environment it sat somewhere else. If they fall away, old addresses lead nowhere and hard-won visibility is lost step by step. The search engine documentation recommends keeping redirects in place for at least one year (Google Search Central) after a move. Language references are affected too: anyone running several language versions should check after the change whether the mutual references still line up, as we describe in our article on structuring and maintaining a multilingual website. The third break point is scheduled tasks: cron jobs for backups, feed imports, invoicing or clean-ups keep running in the old environment until the contract ends, and do not run at all in the new one until somebody sets them up. If you were planning to rebuild anyway, it makes sense to combine the move with a planned website relaunch.

Break pointTypical symptomPrevention before the cutover
CertificateBrowser warning page, forms break offSet up automatic renewal in advance and check it after the change
RedirectsOld addresses lead nowhereExport the rule set, import it again, spot-check the results
Cron jobsBackups, exports and invoices stop happeningDraw up a list with start times and activate them in the new environment
Form deliveryEnquiries stop arrivingSet the sender address and authentication, run a test send
File permissionsUploads fail, preview images are missingAlign owner and permissions after copying
CachesOld content stays visibleClear the caches and set expiry rules again

Email: mailboxes, deliverability and reputation

Email is the part of a move that gets the least attention and does the most damage when it goes wrong. A message that does not arrive produces no error for the recipient, only silence for the sender. Technically the change runs in two steps: first the mailboxes are created in the new environment and the existing messages are synchronised, then the MX record is switched. Between the two steps there is a phase in which messages still arrive in the old environment; a second synchronisation after the switch picks them up. At the same time the authentication markers move along: the SPF record has to reflect the new sending path, the DKIM key has to be installed in the new environment and published in DNS, and the DMARC reports belong under observation during the switch. How these three procedures work together is explained in our article on SPF, DKIM and DMARC.

With the provider, the address you send from usually changes as well. Receiving systems judge that address by its history, and an unknown address starts with no history at all. That is why sending volume should rise gently over several weeks after a change instead of serving the entire list on day one. The requirements of the large mailbox operators set the direction: senders who deliver more than 5,000 messages per day (Google Workspace Help) to the same recipient platform need a published DMARC policy among other things, and the reported spam rate should stay below 0.3 percent (Google Workspace Help) on a lasting basis. Medical practices and other health professions face further requirements on communication and content, which we set out in our article on the practice website and its legal limits.

Dual delivery as a safety net

During the transition phase the old inbox can keep running for a few days and forward automatically to the new one. That way a message lands in the right mailbox even if its sender still has the old MX record cached. After a week the zone has as a rule propagated everywhere, and the forward can be removed. A short test send to several widely used mailbox providers shows before the end of the transition whether authentication and delivery are working cleanly.

Rights to source code and content

A move exposes what was agreed in the contract and what was not. Software is protected by copyright, and rights of use do not arise on their own. Under the purpose-transfer principle in Section 31(5) (UrhG), the scope of granted rights is determined by the purpose assumed in the contract if the types of use are not listed explicitly. In case of doubt the scope stays narrow. For a company that wants to take its website along, that means: if you only agreed on the use of the running site, you do not automatically hold the right to keep operating the source code on another server, to modify it or to have a third party develop it further. Points like these belong in the contract before anything is built, not in a negotiation once the change is already under way.

The same applies to content. Text, photographs and graphics often come from different sources: your own shots, commissioned work, licensed material. Image licences are frequently tied to a particular use and do not travel along by themselves. So clarify before the change which images may continue to be used and for which a licence has to be acquired anew. A clean handover package contains the complete source code, the database, the media files, a list of the extensions in use with their version numbers, the licence records and the credentials. If you would rather hand that responsibility over for good, our website care with a named contact takes it on and keeps the inventory documented as it changes.

The schedule from T-14 to T+7

Two weeks is a realistic frame for a move of medium size. The schedule distributes the work so that the cutover itself is the shortest and best prepared step. With a shop taking live orders, or with connections to inventory management and payment providers, the frame gets longer; for a small company website with no login area it can be shortened. The order of the steps stays the same.

  1. T-14: inventory

    Record contracts, credentials and deadlines. Check the domain holder and technical contact, update the contact details in the register. Choose and commission the target environment.

  2. T-10: backup and build

    Back up the database and directory tree, import them into the new environment and prove the test import. Align extensions, version numbers and server configuration.

  3. T-7: auth code and mailboxes

    Request the auth code, have the transfer lock lifted, transfer the domain. Create the mailboxes and synchronise the existing messages for the first time.

  4. T-2: lower the lifetime and accept

    Lower the lifetime of the affected records to around 300 seconds. Accept the new setup through a local test resolution: forms, login, payment paths, search.

  5. T-0: cut over

    Pause write access briefly, synchronise the database and changed files, switch the zone, change the MX record. Have the certificate issued, check resolution from several networks.

  6. T+1 to T+7: follow-up

    Spot-check redirects, watch the cron jobs, evaluate delivery rates and server logs. Raise the record lifetime again, end the transitional mail forward, cancel the old contract.

A provider change becomes plannable once somebody owns it as a whole instead of splitting it between the old and the new agency. That is exactly how we work: we take the inventory, collect domain, data and mailboxes, build the new setup in a German data centre, agree a fixed cutover time with you and then take over ongoing care and monitoring. What that costs is stated openly in our fixed prices. If a move is coming up, pick the website care option in the enquiry form; we will then ask straight away for the details a workable schedule needs.

Sources and Studies

This article is based on data from DENIC, ICANN, the CA/Browser Forum and the German Federal Statistical Office. Additional material was drawn from Bitkom, the IETF standards RFC 1035 and RFC 2308, Google's documentation on search and mailbox operation as well as the GDPR, the German Copyright Act and recommendations from the BSI. The figures quoted refer to the status at the time of each publication.

Related Articles