Domain IP Lookup

One hostname, several A records

1st answer

93.184.216.34

2nd answer

93.184.216.35

3rd answer

93.184.216.36

The resolver hands back every record, usually in rotating order

Infrastructure & Ops

Why One Domain Can Have Multiple IP Addresses

4 min read

Look up a large site and you often get three, four, or eight addresses back. Look again from another country and the list changes entirely. Four distinct mechanisms produce this, and they have very different implications.

One hostname, several A records

1st answer

93.184.216.34

2nd answer

93.184.216.35

3rd answer

93.184.216.36

The resolver hands back every record, usually in rotating order

One hostname, several A records. The resolver hands back all of them, usually in rotating order.

1. Round-robin DNS

The simplest approach: publish several A records for the same name. Resolvers return them in rotating order, so different clients try different servers first.

Three A records on one name
example.com.    300    IN    A    93.184.216.34
example.com.    300    IN    A    93.184.216.35
example.com.    300    IN    A    93.184.216.36

It distributes load with zero infrastructure. Its limitation is severe, though: DNS has no health checking. If one server dies, its address keeps being handed out until a human removes the record, and roughly a third of visitors get a failure.

2. Geo-routing (GeoDNS)

Here the authoritative nameserver inspects where the query came from and returns a location-appropriate answer. A visitor in Europe gets a Frankfurt address; one in Asia gets Singapore.

This is why two people legitimately get different answers for the same domain at the same moment. Neither is stale or wrong — the nameserver deliberately gave different replies.

Key takeaway

With geo-routing, there is no single correct answer for a domain. The answer is a function of who is asking and from where.

3. Anycast

Anycast is the most elegant of the four and works below DNS entirely. The same address is announced from many locations at once, and internet routing delivers each packet to whichever location is nearest in network terms.

Geo-routingAnycast
Works atThe DNS layerThe routing layer (BGP)
Addresses returnedDifferent per regionAlways identical
Failover speedWaits for DNS TTLSeconds — routing reconverges
Seen in a lookup?Yes — answers differNo — looks like one plain address

1.1.1.1 is the well-known example: the same four digits everywhere on earth, served by hundreds of separate data centres. Anycast is invisible to a DNS lookup, which is precisely why it is so effective.

4. Dual-stack: A and AAAA together

The most common and least interesting reason. A domain publishing both IPv4 and IPv6 returns both, and a client picks whichever it supports.

Not load balancing at all — two address families
example.com.    3600    IN    A       93.184.216.34
example.com.    3600    IN    AAAA    2606:2800:220:1::248

Dual-stack clients generally prefer IPv6 and fall back to IPv4. See A and AAAA records.

What the client does with the list

Getting several addresses back is only half the story — what happens next is decided by the client, and modern behaviour is smarter than it used to be.

The relevant mechanism is Happy Eyeballs, defined in RFC 8305. A browser given both IPv6 and IPv4 addresses does not try one, wait for a timeout, and then try the other. It starts an IPv6 connection, waits a short interval — typically around 250 milliseconds — and if that has not connected, begins an IPv4 attempt in parallel. Whichever completes first wins, and the loser is discarded.

The practical consequence is that a broken address in your list degrades speed rather than causing an outright failure. That is a safety net, not a licence: a dead address still costs every visitor the delay before fallback.

When DNS is the wrong tool for the job

Round-robin DNS is often reached for as a load balancer. It is worth being clear about what it cannot do.

CapabilityRound-robin DNSReal load balancer
Health checkingNone — dead servers stay listedRemoves failed backends in seconds
Traffic weightingNone — roughly evenPer-backend weights
Session affinityNoneSupported
Failover speedWaits for TTL expiryImmediate
CostFreeA service to run or pay for

The honest summary: round-robin DNS distributes load acceptably and handles failure badly. If a server going down must not take a share of your visitors with it, you need health checking — which means a load balancer, an anycast provider, or a managed DNS service that performs health checks and withdraws failing records automatically.

Telling them apart

What you observeMost likely cause
Several IPv4 addresses, same every timeRound-robin DNS
Different addresses from different countriesGeo-routing
One address, fast everywhereAnycast
A mix of dotted and colon addressesDual-stack — not balancing
Addresses reverse to a CDN hostnameCDN edge nodes
Check from different resolvers to spot geo-routing
dig +short example.com @1.1.1.1
dig +short example.com @8.8.8.8

What this means when you look up a domain

It changes how much weight to put on a single answer.

  • Multiple addresses are healthy. They usually indicate redundancy, not misconfiguration.
  • Your answer is not universal. With geo-routing or a CDN, someone elsewhere sees something else, and both are correct.
  • Do not hardcode them. Addresses rotate, get replaced, and are reassigned. Always resolve the name at connection time.
  • Order carries no meaning. Rotation is often deliberate; the first address is not the primary.

The lookup tool lists every A and AAAA record it receives rather than showing only the first, so you can see the complete published set at once — and the bulk mode makes it easy to compare several domains side by side.

Need to check a domain's public DNS addresses? Use the Domain IP Lookup tool.

domain multiple ip addressesround robin dnsanycastdns load balancing

Frequently asked questions