DNS SRV Failover in Practice: Does the Switch Actually Happen When the Resolver Stutters
A failover design is usually drawn as boxes and arrows, with a line running from the primary to the secondary. That line is DNS more often than not, and it is almost never reviewed with the same care as the boxes. A record set published with a long time to live, a resolver that caches aggressively, and a negative answer that sticks around are enough to turn a sound failover into a fifteen-minute outage that nobody can explain from the logs.
What a Service Record Actually Decides
A SIP client does not read a host name and pick an address. It asks for service records, receives a list of targets with priorities and weights, and then decides. Priority determines which group is used. Weight determines the order within a group. The important detail is that a client only moves to the lower-priority group when the higher-priority group is entirely unusable, and unusable is defined at the transport level. A target that resolves, answers, and then rejects the registration is not the same as a target that does not resolve at all.
DNS expresses preference, not health. A failover expressed only in DNS, and tested only by taking a name server down, has been tested against the one failure mode that DNS handles well. The failures that matter in practice are the ones where the primary resolves perfectly and simply will not take the registration.
| Stage | Expected behaviour | How it fails |
|---|---|---|
| Service record lookup | Both priorities returned | Only the primary group is published or cached |
| Address lookup | Address for the selected target | Cached for longer than the service record TTL |
| Transport attempt | Registration accepted | Target answers but refuses the registration |
| Negative answer | Retried after the negative cache expires | Sticks for the full negative caching period |
| Client re-registration | Retried on the configured interval | Interval longer than the recovery target |
Timers Govern How Quickly a Change Is Seen
Two timers control this. The time to live on the service record bounds how long a resolver may reuse an answer. The negative caching timer bounds how long a resolver may reuse a no-such-record answer. Both are set in the zone, but many recursive resolvers apply their own minimums and their own behaviour, and some honour a short time to live on the service record while caching the address records beneath it for far longer.
Recovery time is therefore the sum of several timers nobody owns: the resolver cache, any intermediate cache between the site and the resolver, the client cache, and the re-registration interval on the device. A failover advertised as instantaneous can take minutes to take effect, and the gap between the advertised and the actual behaviour is where an incident gets escalated to a level it does not deserve.
Testing the Switch Rather Than Hoping for It
A failover test is only meaningful if you can observe both ends of the decision. Change the primary so that it stops accepting registrations, then watch what the client does: does it query for new records, does it receive the secondary, does it register there, and how long did each step take. Repeat the exercise with the primary returning an empty answer rather than a timeout, because an empty record set and a refused connection send the client down different paths.
Watch the resolver's own cache statistics during the exercise. A test that passes because the resolver happened to have been refreshed recently proves very little, and it will not reproduce during the incident it was supposed to prevent.
- Publish the secondary in DNS, not only in a design document
- Check the time to live on the address records as well as on the service records
- Decide how negative answers will be cached and test that behaviour directly
- Confirm the client re-registers on an interval consistent with your recovery target
- Test both an empty answer and a refused connection; they behave differently
- Record observed failover intervals and put them in the runbook
- Keep a local resolver entry for the transport host so DNS is not the only path
Design Choices That Make Failover Real
Give the client something to choose between. A secondary that exists only in a slide is not a secondary. Keep the time to live consistent with how fast you can actually change the zone, since a short value combined with a manual update procedure just produces stale answers with a misleading expiry. Make sure the re-registration interval on the device fits the recovery time you are claiming, because no resolver will help a client that has not looked for a change yet.
Where the failure mode is name resolution itself, state the failover target as a fixed address on the device rather than a name. It is an inelegant answer and a completely reliable one when the resolver is the thing that has failed.
Redstone Approach
Redstone platforms support multiple SIP transport targets per trunk, with failover ordered by the platform as well as by name resolution, so the decision does not rest on resolver behaviour alone. Health-check status is reported per target and registration intervals are configurable, which makes a failover test observable during the exercise rather than inferred afterwards from an absence of complaints.
Redstone Systems, Inc. founded in Delaware, USA in December 2002, has been the ODM vendor for many well-known communications companies, serving the Southeast Asian market. In 2020, Redstone Systems will begin to return to the North America market with its self-developed brand.
Redstone has a complete product line of intelligent voice gateways, providing IP-PBXs, analog VoIP gateways (FXS/FXO), digital VoIP gateways (E1/T1), border appliances, and session boundary controllers (SBCs).
With advanced technology in digital signal processor (DSP), speech coding and speech processing, as well as efficient operational tools such as cloud remote management, auto provisioning, Redstone gateways are widely used in markets of enterprise communications, cloud communications, call centers, operators’ IMS/SIP trunks, bringing users friendly, efficient and reliable communication experience.
For more information about Redstone, You can follow us on Facebook, Linkedin, and Youtube to be the first to get the latest news.

Current Location :

