Insights · Technology careers · 6 minute read

Build a DNS Troubleshooting Work Sample That Shows Your Judgment

Show a hiring team how you explain conflicting DNS answers, choose a bounded action and verify the service. Make one small failure reproducible.

By Vitaver Staffing · Published

The mission of a troubleshooting work sample is to let a hiring team see whether you can turn conflicting observations into a verified repair. For a network, support or platform role, a small DNS lab can show that judgment without a large cloud bill or access to an employer’s systems.

Begin with a concrete service question: why can one test client reach a newly added service while another still reports that its name does not exist? Your output should explain the difference, demonstrate the correction and leave another person able to repeat the investigation. This applies The Method and Signal vs. Noise from The Vitaver Playbook: Operating Doctrine & Craft: identify the observation that decides the next step.

Build a small environment with two distinct roles

Use an isolated lab containing an authoritative DNS server for a fictional zone, a separate caching resolver and a simple test service. Configure the resolver explicitly to use the lab’s authoritative server for that zone. Keep both DNS services restricted to the lab; do not expose an open recursive resolver to the internet. Supply versioned configuration and a reset procedure. Use synthetic names and data, with no copied company configuration or credentials.

An authoritative server supplies the zone’s records. A recursive resolver obtains answers for clients and may use cached information. RFC 1034 describes these roles and the distinction between authoritative and cached answers. Your diagram should show which server each client queries, not merely draw a cloud labeled “DNS.”

Before introducing a fault, record the expected hostname, record type, address and application response. Verify that a known existing name resolves through the resolver and that the test service responds. Save commands and observations as text so reviewers can inspect them; screenshots alone make comparison difficult.

Make a cached absence the central experiment

Choose a second name that does not yet exist. Query it through the caching resolver and save the response. Then add its address record to the authoritative zone, use the server’s supported validation and reload procedure, and query that server directly. It should now return the new record. Query the resolver again while its negative cache entry is still valid.

RFC 2308 explains caching of negative answers. For an authoritative negative response, the negative-cache TTL is based on the smaller of the SOA record’s TTL and its MINIMUM field. A resolver can therefore retain a previous answer that a name does not exist after the authoritative data changes. Adding a record does not itself erase that cached answer.

Use deliberately short, documented lab values so the exercise is repeatable in a review session. Record the initial negative response, the authoritative change, remaining negative-cache lifetime and the eventual new answer. Time the steps; if the entry expires before your comparison, reset the lab and explain why that run did not demonstrate the intended condition. Do not invent a result to fit the explanation.

Collect evidence before changing the system

A useful diagnostic record contains the exact fully qualified name, record type, server queried, time, response status, flags, answer and authority sections. ISC’s BIND documentation for dig describes selecting a server and query type. A direct authoritative query with recursion disabled asks a different question from a normal client lookup through its resolver.

Keep those comparisons visible in a short evidence table:

Suggested observations for the lab
CheckWhat it helps establishWhat remains unproved
Query the configured authoritative server directly.Whether that server loaded the new record for the requested name and type.Whether the client uses that server or a resolver still holds an earlier answer.
Repeat through the client’s configured resolver.What that resolver returns now, including a cached negative response.Whether an application has a separate cache or can connect after resolution.
Retry after the observed negative entry expires.Whether this resolver obtains the updated answer in the controlled lab.Whether every production client would behave identically.
Run the test application request.Whether the lab user can complete the intended request.Production availability, security or performance at scale.

A timeout and an NXDOMAIN response require different investigations. If the server gives no reply, first establish reachability and which service is listening. If the direct authoritative query still reports the name absent, inspect the loaded zone and exact name before blaming a recursive cache. Keep each hypothesis beside the observation that would support or reject it.

Show why the tempting shortcut is inadequate

Clearing every cache may make a symptom disappear while destroying the evidence needed to explain it. In this lab, let the negative entry expire and demonstrate the changed answer first. If you also demonstrate a targeted cache clear, label it as a separate experiment, reset to the same starting condition and explain its scope. Do not imply an operator can clear caches belonging to other organizations.

A production owner must choose between waiting, an authorized targeted action or an alternate service route based on impact and available authority. Your portfolio can explain that decision without pretending a lab author has production access. An emergency does not make an untested global DNS change the obvious remedy.

Then complete the application check. A correct DNS answer followed by an application error is evidence to move the investigation to the next layer. State that boundary clearly. Avoid a blanket “wait 24–48 hours” instruction when the evidence identifies a specific cache and an observable expiry.

Make the handoff assessable in one sitting

Package a README with the mission, topology, pinned versions, setup and cleanup, expected observations, actual results and limitations. Include the zone changes and timestamped query outputs. Explain your personal contribution and credit any borrowed configuration or AI assistance. Ask a peer to reset the lab and reproduce the mismatch and recovery without coaching. Correct the first instruction they cannot follow.

Keep the demonstrated scope honest. This sample does not establish skill in public delegation changes, DNSSEC validation, split-horizon production networks or large-scale incident response. Name one extension that would answer a relevant next question rather than adding unrelated tools. For example, a second lab client could show that different resolvers have different cache histories.

Connect the evidence to a real hiring decision

A recruiter should be able to send a concise note: the candidate built the lab, distinguished authoritative data from a cached negative response, and supplied a repeatable application check. Keep that demonstrated ability separate from unverified claims of production experience. Confirm the candidate’s actual site, shift and support availability for the assignment.

If a buyer uses this as an assessment, provide the same starting environment, permitted tools, effort limit and job-related criteria to everyone assessed for the role. Use fictional infrastructure rather than ask applicants to repair the company’s live service. The technical evaluator should score evidence selection, diagnosis, bounded action and completion checking. The client’s hiring owner can then decide whether the demonstrated troubleshooting is sufficient for the actual responsibility or whether a specific further skill needs assessment.

Questions for discussion

  1. Which observation would make you stop blaming a DNS cache and investigate the authoritative server?
  2. When does clearing a cache solve the operating need, and what evidence should you preserve first?
← All Insights