What is DNS monitoring? Records, resolvers, and alerts

Learn how DNS monitoring checks availability, lookup time, and expected records, and how caches, CDNs, and DNSSEC change the way you interpret failures.

Best practices
A conceptual illustration of services connected to monitoring locations across regions

DNS monitoring regularly checks whether DNS queries succeed, how long they take, and whether the answers match the service's intended configuration. It helps separate a name-resolution problem from an application failure. A healthy server is of little use to a visitor whose resolver returns the wrong address.

Address records are only part of that job. DNS also publishes aliases, mail routes, authoritative-server information, and policy records. An unexpected change deserves investigation, but a changed IP address does not by itself prove hijacking. The first task is to define which answers are valid for the name, record type, and location being checked.

Start with the query and its expected answer

A monitor sends a query on a schedule and applies rules to the result. A manual lookup might start with:

dig example.com A

That command asks for IPv4 address records through the resolver selected by the local environment. It does not compare the result with your deployment inventory. The monitor needs a separate expectation: an approved address set, a permitted alias target, a response-time threshold, or a required record.

Suppose a fixed-address service is expected to return 192.0.2.10. An address outside its approved set can trigger an alert. This is a hypothetical example: 192.0.2.10 belongs to a documentation range and is not the actual fixed public address of example.com. A CDN or geographically routed service needs a rule that allows its legitimate answers.

Record query errors and timeouts as well as mismatches. A failed DNS lookup can stop an HTTP request before it reaches the application, while an answer that looks syntactically valid may still send the request to the wrong endpoint. Those failures require different investigations.

Check the records the service actually depends on

Record What it describes What to check
A / AAAA IPv4 / IPv6 addresses Whether the returned addresses belong to the service's approved configuration
CNAME An alias to another name Whether the target and the relevant alias chain are intended
MX Mail servers and their preferences Whether delivery still points to the approved servers
NS Authoritative name servers Whether delegation matches the approved DNS-provider configuration
SOA Zone version and timing information Whether expected zone-version changes reach the authoritative servers
TXT Text records used by verification and policies such as SPF and DKIM Whether required values remain present and correct

An MX change can break delivery or direct mail elsewhere, but a missing MX record does not always mean that the domain cannot receive mail. SMTP defines an implicit-MX fallback to the domain's address records when no MX records are present. An explicit null MX has a different meaning: the domain declares that it does not accept mail. SMTP routing · Null MX

A changed NS set may be an approved provider migration. An unexpected delegation is more concerning, but it still needs to be checked against the change history. Likewise, the SOA serial is a zone-version value, not a universal audit trail. It does not have to increase by exactly one, and polling it cannot establish that every DNS record is correct. SOA serial arithmetic

Compare locations without assuming identical answers

DNS is distributed. A recursive resolver may answer from its cache, and an authoritative service may select an answer based on location or client-subnet information. Internal and public resolvers can also be configured to expose different views. Consequently, two different IP answers can both be correct.

Use monitoring locations that represent the paths your users rely on. Where reachable, public resolvers such as 8.8.8.8, 1.1.1.1, and 223.5.5.5 provide additional points of comparison. They are useful observations, not a vote in which the majority automatically identifies the correct answer. Compare differences with the authoritative configuration, cache state, and intended routing policy. Google Public DNS ECS guidance

This is where Guance Synthetic Monitoring can help narrow the investigation. Selected public or self-built locations provide outside-in timing, including DNS lookup, connection, and time to first byte. A failure concentrated in the DNS phase calls for a different next step from a slow server response. For MX, NS, or SOA content checks, use a separate DNS checking tool or script with verified support for the required queries and assertions. Recording the DNS phase of an HTTP check does not validate those records.

Read TTL as cache behavior

After a record changes, a recursive resolver can continue serving its previously cached answer. The remaining TTL is relevant, but it is not a countdown to guaranteed worldwide completion. Negative answers can be cached too, and delegations and alias chains have their own cached records.

Some resolvers also support serving stale data when they cannot refresh it from authoritative servers. That can preserve availability during an upstream DNS failure while making the returned answer older than its original TTL. Check authoritative answers and representative recursive paths separately. A successful result at your chosen locations describes those observations; it does not prove that every user now sees the new configuration. Negative caching · Serving stale data

Follow the failed dependency

Several incidents can look like the same unavailable website:

  • An unauthorized address change can direct visitors to another service. A monitor can identify the deviation only on paths it covers and with an assertion that detects it.
  • An accidental change can send the production name to a test endpoint.
  • An authoritative DNS outage may cause fresh lookups to fail while cached answers keep other clients working.
  • After an address change, the new HTTPS endpoint may lack a valid certificate for the requested name. Changing the IP itself does not change the certificate's hostname requirement.

Keep DNS checks beside HTTP and TLS checks. A valid DNS answer cannot prove that the destination serves the right application, and a working application at one location cannot rule out a resolver problem elsewhere.

Use rules that fit CDNs and the monitoring budget

For a CDN, an exact-IP rule is often too narrow. Check approved CNAME targets and expected routing behavior, then verify HTTP and TLS behavior at the destination. A provider's published and maintained IP ranges may support an additional rule, but an address belonging to that provider does not alone prove that it serves your application.

For a critical domain, a one-to-five-minute interval can be a starting point, adjusted to the detection delay the business can tolerate. A single query is small; the total number of names, locations, and scheduled runs still affects query load and monitoring charges. Choose the coverage deliberately rather than assuming high-frequency checks have no cost.

DNSSEC and monitoring answer different questions

With a valid trust chain and validation enabled, DNSSEC authenticates signed DNS data and protects its integrity. It does not encrypt queries, and a valid signature cannot establish that an authorized configuration matches the business's intent. A mistakenly published address can still be correctly signed.

Monitoring checks reachability and expected results along the paths it observes. It can identify some unintended changes even without DNSSEC, but it cannot guarantee detection of every form of tampering. Use DNSSEC validation, configuration controls, and monitoring together. DNSSEC capabilities and limits

Get a tailored plan

Contact us

Join the community

Scan with WeChat
to join the community

Try Guance

Start online and pay only for what you use.

Get started

Choose a Guance plan

Code hosting