The DNS layer is the most often misdesigned piece of a cPanel architecture. Default cPanel DNS — authoritative BIND on the WHM box, exposed to the internet on port 53 — works at small scale but fails badly under DDoS, doesn’t survive a single host outage, and offers no geographic redundancy. The fix is a cluster, but the shape matters.
The trade-off
Three common topologies:
- Standalone authoritative — DNS lives on the WHM host. Simplest; single point of failure for both DNS and DDoS surface.
- Hidden primary + N public secondaries — the WHM host is the primary but not exposed; named slaves (DNSONLY hosts) sit in the public DNS path. Primary outage doesn’t affect resolution. Standard cPanel pattern.
- Hidden primary + external anycast secondaries — primary is your WHM cluster; secondaries are a paid anycast service (Cloudflare DNS, AWS Route 53, DNSimple, NS1). Best DDoS posture; highest cost.
How to decide
- Fleet under 10 hosts, low risk profile? Standalone or basic primary-secondary with two DNSONLY nodes. Don’t overcomplicate.
- Customer-facing brand with reputation cost on DNS outage? Hidden primary plus anycast secondaries. The cost of a DNS outage exceeds the cost of the anycast service.
- Compliance regime requires DNS query logging? Be explicit — anycast providers log differently than self-hosted BIND. Match the logging surface to what the auditor needs.
- Mail-heavy workload? DNS for MX/SPF/DKIM/DMARC must resolve cleanly from anywhere. Anycast or geographically distributed secondaries help here.
- Multi-region cPanel cluster? Authoritative DNS should not be co-located only in one region. Either multi-region secondaries or anycast.
What to do
- Default to hidden primary + at least 2 DNSONLY public secondaries on different /24s and ideally different IaaS providers.
- Set NS records to point at the secondaries, not the primary. The primary should not appear in
dig NS yourdomain.com. - Configure
notify-sourceandtransfer-sourcecorrectly so secondaries pull from the right interface. - Test failover. Stop named on one DNSONLY node and confirm queries continue answering from the other. Stop named on the primary and confirm zone updates pause but resolution continues.
- Monitor zone transfer lag. WHM > Clusters shows it; if it’s growing, the secondary isn’t keeping up.
- If you adopt anycast secondaries, push zones via the provider’s API or AXFR depending on what they support. Most anycast providers accept zone push from a hidden primary.
- For an example of the anycast layer, see BGP anycast BIRD2 deploy — the underlying pattern even if you don’t run BIRD yourself.
Stack Harbor designs DNS cluster topology per fleet as part of managed cPanel operations.