Engineering Glossary · Addressing
IPv6 Is Invisible Until an AAAA Record Points at Nothing
A handful of visitors report the site hangs for a few seconds and then loads, and it works perfectly from your machine.
The short answer
IPv6 is the addressing scheme with an effectively bottomless supply of addresses, running alongside the exhausted IPv4 internet. AAAA records map hostnames onto it, and dual-stack servers answer over either protocol.
It stays invisible right up until an AAAA record points somewhere that no longer answers. Browsers implement Happy Eyeballs — they race a v6 and a v4 connection and use whichever completes first — so a broken v6 path usually costs a small delay rather than an outage, which is precisely why it can persist for months without anyone diagnosing it.
By the HostingFast team · Reviewed 12 August 2026
0
Entries with no jargon left over
100+
Terms, all cross-linked
Real
Worked examples, not diagrams
Free
To read, permanently
The long hexadecimal addresses end scarcity permanently, and the transition happens by coexistence rather than by a switch-off date. Decades of hardware and software assume IPv4, so both run side by side and will for a long while yet.
From a site owner's chair the whole thing is normally automatic. Turn dual-stack on and a visitor whose connection speaks IPv6 is served over it without a decision being taken at either end.
Testing both paths separately
Force each protocol and compare. curl -4 against your URL and curl -6 against the same URL will tell you in two commands whether both stacks serve the site, and whether one of them is slower or simply does not answer.
Do the same for the record itself: query for A and for AAAA and check both point where you expect. A leftover AAAA from a previous host is one of the tidiest ways to give part of your audience a worse experience than the rest.
Happy Eyeballs, and why the fault hides
Browsers do not commit to one protocol. They start connections over both and keep whichever succeeds first, falling back quickly when one path stalls. The result is that a broken v6 route produces a short, intermittent delay rather than a visible failure.
Intermittent and small is the hardest class of bug to get reported accurately. The user says it is sometimes slow, your own testing shows nothing, and the actual cause is an address in a record nobody has looked at since the last migration.
Where IPv6 does need a decision
Two places. If you maintain a zone by hand, publishing an AAAA record is a deliberate act and so is removing one. And if you administer a VPS yourself, enabling IPv6 and configuring the firewall for it are jobs nobody else will do on your behalf.
That second one deserves emphasis: firewall rules written only for IPv4 leave the v6 stack unfiltered on a dual-stack box. Rules have to exist for both address families or the protection is only half present.
Dual-stack behaviour on this platform
Where dual-stack is enabled the platform serves over either protocol without configuration on your part, and free SSL covers the hostname regardless of which address a visitor arrived on.
On a VPS you hold root and therefore the decision. KVM virtualisation, a dedicated IPv4 and DDoS filtering come with the plan; what you do with the v6 stack and its firewall rules is yours to configure.
The bugs that hide behind a fallback
The most expensive faults are the ones that mostly work. These entries pay particular attention to the failures that degrade quietly instead of breaking loudly.
NVMe storage and server-level LiteSpeed caching are on every tier, from the smallest plan upwards, rather than reserved for the expensive end of the range.
- 100+ entries, quiet failures included
- Both protocols tested, not assumed
- Wired into IP Address, A Record and DNS
- Written by the engineers who chase these
Why HostingFast
Standard on every plan
Both stacks, tested
Two commands prove whether v4 and v6 both serve your site, and how fast each one is.
IPv6, as a failure mode
The entry is organised around the intermittent slowness a stale AAAA record produces.
Happy Eyeballs explained
Why a broken path shows up as a delay rather than an error, and why that makes it survive.
Firewall warning
Rules written for one address family leave the other unfiltered on a dual-stack machine.
Where a decision is required
Hand-maintained zones and self-administered servers — the only two places it is your call.
Continues into its neighbours
IP Address, A Record and DNS carry on from here.
Quick Start
From order to online
- 1
Force each protocol in turn
Request your URL over IPv4 only and then over IPv6 only. Compare the results and the timings before you conclude anything.
- 2
Check the AAAA record matches reality
Query for AAAA and confirm the address still belongs to your current server. Stale records survive migrations remarkably well.
- 3
Remove a v6 address you cannot serve
If the v6 path does not answer, delete the AAAA record. A missing record is faster than a broken one every single time.
- 4
Mirror your firewall rules
On a server you administer, write rules for both address families. One-family rules are a half-configured firewall.
Built In
Loaded onto every plan
- Dual-stack serving where enabled, with no configuration required
- Free SSL covering the hostname on either protocol
- Full A and AAAA record editing in the panel
- KVM virtualisation with root access on VPS plans
- Dedicated IPv4 included with every VPS
- DDoS scrubbing handled out at the network edge
- SSH, Git and Composer on developer-focused plans
- In-place plan upgrades with no migration
Frequently Asked
What people ask us most often
Some visitors say the site is intermittently slow. Could IPv6 be the cause?
It is a strong candidate. If an AAAA record points at an address that no longer answers, IPv6-capable visitors attempt that path first, stall, and fall back to IPv4 — producing a short delay for some people and nothing at all for everybody else. Query the AAAA record and request the site over IPv6 explicitly; two commands will settle it.
What is Happy Eyeballs actually doing?
Racing the two protocols. The browser starts connections over IPv6 and IPv4 more or less together and keeps whichever completes first, abandoning the other. It is what makes a broken v6 path degrade rather than fail, which is excellent for users and unhelpful for anyone trying to notice the fault.
Do I need to configure IPv6 on my hosting?
Almost never on a shared account — dual-stack is enabled where it is available and there is nothing to switch on. It becomes your job in two situations: publishing AAAA records in a zone you maintain yourself, and enabling and firewalling IPv6 on a server where you hold root.
What happens when I outgrow my plan?
You upgrade in place. Plan changes happen from the client area with no migration and no downtime, and the range runs from compact shared accounts through VPS to full dedicated servers, so growth is an account setting rather than a hunt for a new host.
Keep reading
A Record
The IPv4 half of the pair, and the record most cutovers actually change.
IP Address
The address as a latency budget, and how to split network time from server time.
Best LiteSpeed Hosting
The same call made as a buyer's guide: what to verify and what to ignore.
WooCommerce Hosting
Store-ready hosting with Redis caching and headroom reserved for checkout traffic.
VPS Hosting
KVM virtual servers with root access, DDoS filtering and a flat monthly price.
Changing hosts? Run through our checklist first.
A straightforward sequence for a switch your visitors never feel: which files move first, how to shift email across without losing a single message, the right moment to repoint DNS, and the two mistakes behind almost all the downtime we get asked to rescue.
Serve both paths properly.
NVMe, LiteSpeed and free SSL on every plan, engineers who answer at any hour, and a renewal price that matches the day you signed up.
View WooCommerce Hosting plans