Why Is My Domain Still Pointing to the Old Server? DNS Caching and TTL Explained
Imagine moving example.com to an AlexHost VPS. You update the domain’s address, then check the website. Your phone on mobile data shows the new site. Your laptop on home Wi-Fi still shows the old one. Did the change work?

In this hypothetical move, the address changes from 192.0.2.10 to 203.0.113.20. Both are documentation-only examples, not real hosting infrastructure or destinations to visit. Publishing a new DNS address does not erase earlier saved answers. Some visitors may receive the old address until those answers expire.
But that is only one possibility. Incorrect DNS configuration can keep publishing the wrong address, and a correct lookup can still lead to old website content. The phone/laptop disagreement is a clue, not a diagnosis. Understanding DNS caching and TTL helps you decide whether to wait or investigate—without terminal commands.
Where your browser gets the domain’s address
The Domain Name System (DNS) holds named entries called records. For this website, an A record supplies an IPv4 address—the dotted-number format shown above. Other record types hold different information.
An authoritative nameserver publishes the domain’s records. A recursive resolver looks up answers for your device and may reuse a cache: a temporary saved answer. Caching avoids repeated work and speeds up later lookups. Cloudflare’s DNS overview explains these roles.
LOOK UP THE ADDRESS
Browser
↓ if no saved answer
Device's DNS handling
↓ if no saved answer
Recursive resolver
↓ if no saved answer
Authoritative DNS → published answer returns to the browser
========
THEN REQUEST THE WEBSITE
Browser → resulting server/proxy address → page response
The key distinction is between the published answer and saved copies. Our example keeps the same authoritative DNS provider and nameservers: the edit changes the website’s address, not who publishes it. TTL explains how long an earlier copy can remain in use.
TTL explained: the timer belongs to the cached answer

Time to Live (TTL) normally limits how long a saved DNS answer may be reused before it needs refreshing from its source. It is measured in seconds:
- 300 seconds = 5 minutes;
- 3600 seconds = 1 hour;
- 86400 seconds = 24 hours;
It is not the time an edit takes to save or your website’s lifespan.
Consider two resolvers obtaining the old address five minutes apart, each with a one-hour TTL. At 12:10, authoritative DNS changes the address and publishes a shorter TTL:
OLD ADDRESS: 192.0.2.10 — TTL 3600 seconds (1 hour)
Resolver A: saved at 12:00 → original expiry about 13:00
Resolver B: saved at 12:05 → original expiry about 13:05
12:10 — Authoritative DNS publishes 203.0.113.20, TTL 300.
12:15 — A resolver with no saved answer can get the new IP.
Both servers can therefore receive visitors during the transition. A successful early check shows that one lookup found the new address, not that traffic to the old server has stopped. Expiry does not schedule a query: refresh normally happens when an answer is next needed, though some resolvers refresh early.
📝 Note: Reducing the published TTL affects subsequently obtained answers. It does not shorten countdowns already issued with the old address, even when the new value is only five minutes.
Think of an official address directory as authoritative DNS. Assistants act as resolvers, keeping timestamped notes so they can reuse an address without consulting the directory again. Those notes are the cached answers.
A note marked “use for one hour” remains usable for that interval even if the directory changes. No recall notice erases the notes. In real DNS, timers and refresh rules govern that reuse.
Ordinary downstream DNS caches receive the remaining TTL, not a restarted full lifetime: passing an answer along does not grant another full hour. RFC 1034 illustrates this cache aging. Browser and application caches may follow different policies, so the DNS countdown does not describe every layer’s behavior.
Why visitors disagree—and what “DNS propagation” really means
Different networks can use caches populated at different times. Some never held the old answer; others share a resolver serving many visitors. Physical distance from the new server is not a simple countdown. In our unchanged-DNS-provider example, DNS propagation commonly describes earlier cached answers gradually being replaced—not a worldwide push that reaches every device in order.

Changing nameservers is a separate operation. Delegation identifies a domain’s authoritative nameservers, and that nameserver (NS) information has its own caching lifetime. Lowering the website’s A-record TTL does not shorten those delegation timers. A website move need not involve changing DNS providers or transferring registration; separating those changes makes the timing easier to understand.
There is no universal 24-, 48-, or 72-hour rule. The previous relevant TTL gives a normal-behavior estimate, not an end-to-end deadline. Clicking Save may also precede publication: the provider needs to make the update available on the relevant authoritative servers. As an exception to ordinary expiry, some resolvers can serve expired data when attempts to refresh fail, a behavior called serve-stale.
When waiting will not fix it: identify the layer
Compare the published answer with the affected lookup
Start with the exact hostname—the name being looked up. example.com is the bare/root domain; www.example.com is a separate hostname. Compare the same name and record type across the relevant active authoritative servers, not just the account you edited.
This is where provider roles matter. The registrar manages registration; the web host serves the site. The authoritative DNS provider publishes the domain’s records, grouped in a zone. One company can fill several roles, but an edit matters only if it reaches the active zone served by the designated nameservers.

You can make an initial comparison in your browser with Dig Web Interface. Use your own affected hostname—not the documentation example—and keep the record type the same for both lookups:
- Enter the hostname under Hostnames or IP addresses and choose A under Type. Under Nameservers, select Authoritative, then click Dig. Review the address returned by each listed authoritative server; leave Only first nameserver unchecked.
- Select Resolver and choose a service such as Google from the dropdown. Click Dig again to compare that resolver’s answer with the published answers. This queries the selected service from the tool; it does not change your device’s DNS settings.
In our hypothetical move, authoritative answers of 203.0.113.20 alongside a resolver answer of 192.0.2.10 would point toward an earlier cached copy at that resolver. If authoritative servers still return the old address, investigate publication or configuration instead. Compare your own old and new addresses with this pattern.
The comparison follows Google’s DNS troubleshooting guidance: querying authoritative DNS directly avoids a recursive resolver’s saved answer for that record. Returned TTLs, where shown, add timing context; record when you made the comparison. After a recent nameserver change, ask the DNS provider to help compare both earlier and current delegation paths, rather than relying only on automatic server discovery.
The evidence separates into three branches:
- Authoritative answers are wrong or unexpectedly disagree: investigate active-zone configuration or publication. Waiting for recursive caches will not repair the underlying fault.
- Authoritative answers are correct; the affected lookup returns the earlier value: caching is plausible. Interpret its remaining lifetime alongside the earlier TTL.
- The intended public endpoint is reached; old content remains: investigate proxy origin selection, redirects, content caches, or the new server’s site configuration.
This browser check is a starting point, not a test of your laptop’s exact path: the browser itself may use a separate secure-DNS resolver. Multi-location checkers likewise sample paths rather than measure a worldwide “percentage propagated.”
Match the evidence to the layer that needs attention

An A record supplies IPv4; an AAAA record supplies IPv6, another Internet address format. A CNAME instead aliases a name to another hostname. That target must resolve correctly, but the alias itself is not an HTTP redirect.
A content delivery network (CDN) or proxy sits in front of an origin, the server behind it. Our direct-to-server example expects the VPS address. With a proxy, the correct public answer may belong to the intermediary instead. The following table connects these differences to evidence and the layer that needs attention:
| Possible cause | Evidence that points to it | What needs attention |
|---|---|---|
| Wrong or inactive DNS-management zone | Your dashboard shows the edit, but active authoritative DNS publishes another value. | Confirm the edit was made in the zone served by the domain’s designated nameservers. |
| Inconsistent authoritative publication | Relevant servers unexpectedly disagree for the same name/type in this static-address setup. | Ask the DNS provider to investigate publication, replication, or configuration. |
| Old AAAA or unintended extra address | A looks correct, but visitors using IPv6 or an extra returned address still reach the old server. | Review the complete A/AAAA set and intended service; correct the mismatch, not IPv6 generally. |
| Bare-domain, www, or CNAME-target mismatch | example.com resolves as intended; www.example.com or its alias target resolves elsewhere. | Review that hostname and alias target; changing the bare domain does not update unrelated entries. |
| Wrong proxy/CDN origin | Public DNS correctly reaches the proxy, but its backend setting still selects the old server. | Review which origin is configured to handle requests for this hostname. |
| Content cache, redirect, or new-server site configuration | The intended endpoint is reached, but delivery evidence identifies a saved page, redirect, or wrong site. | Investigate the implicated browser/CDN/application cache, redirect destination, or new-server site mapping. |
DNS neither performs HTTP redirects nor stores web pages. To identify which server delivered a page, pair DNS results with delivery evidence, such as server-side request records. Multiple addresses can be intentional; an address inconsistent with your deployment is what warrants investigation. If you cannot access that evidence, ask the DNS or hosting operator to help identify the layer.
A related case: when DNS remembers a missing name
If a newly created app.example.com appears missing rather than reaching an old server, the remembered answer may be NXDOMAIN: “the queried name does not exist.” This is negative DNS caching. A name can also exist without the requested record type, so first check what authoritative DNS publishes.
The negative answer’s lifetime comes from resolver policy and the zone’s Start of Authority (SOA) record, which carries zone-level settings—not the new A record’s TTL. Lowering that new TTL cannot recall the earlier negative answer. As RFC 2308 explains, ordinary cache-hit queries do not restart its countdown.
Will refreshing the page or clearing a cache help?
Saved page content is separate from saved DNS answers. Those DNS answers may be held locally by your browser or device, or upstream by a resolver. A hard refresh may help with page content, but it is not a reliable DNS-cache flush. Private browsing does not prove that a fresh DNS lookup occurred either.
Local handling can include an operating-system cache or an optional router forwarding queries. Upstream, a resolver run by your internet service provider or a public service can reuse an answer for many users. That difference in scope explains why a local action may change one visitor’s result without affecting others.
Clearing a local DNS cache can help when that cache holds the earlier answer, but it does not clear an upstream resolver. Another network or resolver offers comparison evidence, not a global fix. Some public resolver services offer purge tools; those affect only that service’s relevant cache entries, not everyone’s answers.

When asking your DNS or hosting provider for help, include these details where available:
- Affected domain and exact hostname, plus the record type being compared.
- Change time and timezone, with the previous TTL if you know it.
- Expected and observed answers, including when each observation was made.
- Resolver/server or network used, plus any available comparison with authoritative answers.
Make the next server move less surprising

For the next move, work backward from cutover: the moment you switch public DNS to the new address. Use three phases to prepare for visitors reaching either server:
- Before cutover: Inventory the website’s relevant A, AAAA, and CNAME paths, and confirm the new website is ready. Where supported, publish a temporary lower TTL in advance. If the previous TTL was 86400 seconds, wait at least 24 hours after the relevant authoritative servers publish the lower value, under normal behavior, before switching the IP. This lets previously issued longer lifetimes expire. A temporary value such as 300 seconds is illustrative, not mandatory or universally supported.
⚠️ Warning: Orders, uploads, and database writes need a coordinated migration plan. Two independently writable copies can split or lose data. Keeping both servers online is not, by itself, a safe transition strategy for a dynamic website.
- During and after cutover: Keep a safe, working old path available while earlier caches turn over. Monitor relevant DNS answers alongside traffic to the old and new servers. Use the earlier TTLs and observed behavior to choose an overlap period, allowing an appropriate margin rather than a universal number of hours. Avoid repeated speculative edits during that period.
- Once stable: Return applicable TTLs to a value appropriate for the service. Shorter TTLs allow quicker adaptation but generally create more DNS queries and cache misses. Longer ones improve reuse but lengthen the ordinary change window.
If you also change DNS providers, use a dedicated migration plan. Keep the zones consistent during the transition, and include DNSSEC planning where relevant. AlexHost’s DNS service configuration guide covers setup if you separately choose to use that service; buying a VPS does not activate it automatically.
The bottom line: check the answer before blaming propagation
Return to the phone showing the new site and the laptop showing the old one. Three questions now guide the next check:
- What do the active authoritative servers publish?
- What does the affected lookup path return?
- Is the intended endpoint delivering old content?
The answers distinguish a countdown from a DNS configuration or website-delivery problem. Time can retire an earlier cached answer. It cannot correct an inactive zone, a forgotten address, or a wrong origin. Wait when the evidence points to caching; otherwise, investigate the layer it identifies.
on All Hosting Services