Back in September we wrote about how we built our clustered DNS infrastructure - two NSD nodes across Sydney and Melbourne, self-healing peers, and a reconciliation loop that keeps them in sync without manual intervention.
We ended that post with a note: “We’re also planning a tertiary node closer to home.” Closer turned out to be an understatement. That node is now live and it’s in New Zealand!
ns3: On Home Soil
The third node brings the cluster to three geographically separated nameservers - and unlike the existing Australian nodes, ns3 is hosted right here in New Zealand, on local ISP infrastructure. All three run the same NSD stack, participate in the same peer reconciliation loop, and serve the same 260+ zones.
The practical effect is that a DNS lookup for any domain we host now has three independent places it can be answered - and the one that responds quickest wins. For visitors querying from New Zealand, that now means a server that doesn’t need to cross the Tasman at all. A DNS response from local infrastructure is measurably faster, and for a lookup that happens before your browser even tries to load a page, every millisecond counts.
Why Three Matters
Two nodes gives you redundancy. Three gives you something more useful: majority consensus.
With two nodes, if they ever disagree on zone state - one has a recent update, one doesn’t - there’s no tiebreaker. Each node thinks it’s right. With three nodes running the same reconciliation logic, the odd one out gets corrected by its two peers before it has any meaningful opportunity to serve stale data.
It also means the cluster survives a node loss without any reduction in geographic coverage. Sydney can go offline and you still have Melbourne and the new node answering queries. Melbourne can go offline and nothing changes for end users. The cluster keeps serving without degrading.
What Hasn’t Changed
The architecture is the same as described in the original post. DirectAdmin still acts as the zone data source, pushing updates to all nodes simultaneously. Each node still runs the peer reconciliation loop, still checks back against the DA master hourly as a safety net.
The new node joined the same way any node would: it pulled the full zone set from an existing peer on first sync, then started participating in the regular reconciliation cycle. No special bootstrap procedure, no manual zone transfers. The self-healing design meant adding a third node was straightforward.
The Bigger Picture
NSD’s minimal footprint - it uses a fraction of the memory of alternatives we evaluated - makes this practical to run on distributed infrastructure without significant ongoing cost. That efficiency is part of what lets us invest in proper geographic redundancy rather than cutting corners on a single node and hoping for the best.
It’s not the kind of infrastructure most small NZ hosts think about. For most providers, DNS is an afterthought bolted on by the registrar. For us it’s a first-class part of the stack - because we’ve seen what happens to sites when it isn’t.
Three nodes. Three regions. 260 zones. All of it reconciling itself automatically while we sleep.
Hosting built on infrastructure like this, starting at $6/month. View our plans or get in touch if you want to know more.