AAAA Records: Mapping Names to IPv6 Addresses

The Internet ran out of addresses.

Not metaphorically. Literally. IPv4 provides about 4.3 billion addresses, and we used them all. Every smartphone, every smart thermostat, every security camera—they all need addresses, and 4.3 billion wasn't enough for a world with 8 billion people and dozens of devices each.

AAAA records are DNS's answer to this crisis. They map domain names to IPv6 addresses, just as A records map names to IPv4. The name tells you exactly what happened: we needed four times more space. IPv6 addresses are 128 bits versus IPv4's 32 bits—four times larger—so the record type got four A's instead of one.

What an AAAA Record Looks Like

example.com.    3600    IN    AAAA    2001:0db8:85a3:0000:0000:8a2e:0370:7334

This says: "example.com lives at this IPv6 address. Cache this answer for 3600 seconds."

IPv6 addresses look different from IPv4. Instead of four decimal numbers separated by dots (192.0.2.1), you get eight groups of hexadecimal separated by colons. They're longer, but they can be shortened—consecutive groups of zeros collapse into ::. So 2001:0db8:0000:0000:0000:0000:0000:0001 becomes 2001:db8::1.

The mechanics are identical to A records. Client asks DNS, DNS returns address, client connects. The only difference is the size of the address being returned.

Living in Two Worlds

The Internet can't switch from IPv4 to IPv6 overnight. Billions of devices, decades of infrastructure, countless systems that only speak IPv4—you can't just turn them off.

So we run both. This is called dual-stack:

example.com.    3600    IN    A       192.0.2.1
example.com.    3600    IN    AAAA    2001:db8::1

Same domain, two addresses. IPv4 clients use the A record. IPv6 clients use the AAAA record. Everyone gets in.

For high-traffic services, you'll see multiple records of each type:

example.com.    300     IN    A       192.0.2.1
example.com.    300     IN    A       192.0.2.2
example.com.    300     IN    AAAA    2001:db8::1
example.com.    300     IN    AAAA    2001:db8::2

This provides redundancy in both protocol families. If one server fails, traffic shifts to another. If IPv6 breaks, IPv4 still works.

Happy Eyeballs

When your browser sees both A and AAAA records, how does it choose?

It races them.

The algorithm is called Happy Eyeballs (yes, really—that's in the RFC). Your browser starts connecting via IPv6, then 50-250 milliseconds later starts an IPv4 connection too. Whichever completes first wins. The loser gets dropped.

This sounds wasteful, but it's actually clever. IPv6 gets a head start, so it wins when both work equally well—encouraging adoption. But if your IPv6 is broken or slow, IPv4 wins automatically. You don't wait for a timeout. You don't configure anything. You just get the fastest working connection.

The algorithm also remembers. If IPv6 failed last time, it adjusts the head start. If a destination always works better on IPv4, it learns. Your connections get faster over time.

The Transition Is Happening

Over 40% of traffic to Google now arrives via IPv6. Some countries exceed 70%. Mobile carriers lead the charge—it's easier to deploy IPv6-only cellular networks than to manage dwindling IPv4 allocations.

This matters because some networks are going IPv6-only. They use translation technologies (NAT64/DNS64) to reach IPv4-only servers, but that translation adds latency and complexity. Services with AAAA records bypass the translation entirely.

If you don't publish AAAA records, you're not just missing IPv6 users—you're making things slower for users on transitional networks. As more carriers go IPv6-only, the cost of not having AAAA records grows.

Getting It Right

Only publish AAAA records for working IPv6. This sounds obvious, but it's a common mistake. If your server doesn't actually have IPv6 connectivity, publishing an AAAA record just causes failures. Happy Eyeballs will eventually fall back to IPv4, but you've added delay and broken the experience for IPv6-only clients.

Match your TTLs. If your A record has a 3600-second TTL and your AAAA record has 300 seconds, caches will have one address but not the other. This creates inconsistent behavior and unnecessary DNS queries.

Test from IPv6-only networks. The easiest way to find IPv6 problems is to remove IPv4 as a fallback. Tools like test-ipv6.com help, but nothing beats testing from an actual IPv6-only connection.

Monitor both protocols separately. IPv6 traffic patterns differ from IPv4. You might have great IPv4 performance and broken IPv6, or vice versa. If you're not measuring both, you're flying blind.

The Bigger Picture

AAAA records exist because the Internet outgrew its original addressing scheme. They're the DNS component of a massive, ongoing infrastructure migration—the largest coordinated change in Internet history.

The dual-stack era won't last forever. Eventually, IPv4 will fade the way dial-up faded. When that happens, services without AAAA records will simply be unreachable to a growing portion of the Internet.

That future isn't hypothetical. It's already happening on mobile networks. It's already happening in data centers that can't justify the cost of IPv4 addresses. Every year, the Internet becomes a little more IPv6.

AAAA records are how your services stay findable in that future.