DNS is a cached, hierarchical lookup; most DNS incidents are really caching (TTL) incidents.
01 · Why it’s true
Names are resolved by asking a chain of servers. Your machine asks a recursive resolver; if the resolver has no cached answer, it asks a root server, which points it to the servers for the top-level domain (like .com), which point it to the domain’s authoritative servers, which return the answer.
Every answer carries a TTL (time to live) that tells resolvers how long they may cache it. Caching is what makes DNS fast and scalable — and it’s why DNS changes don’t take effect instantly everywhere, and why a bad record can linger after you’ve fixed it.
02 · Picture it
A cache miss walks the hierarchy in 8 steps; within the TTL, the resolver answers in 2.
03 · In the interview
Explaining why a DNS change “didn’t work” for some users: they still hold a cached answer.
Choosing TTLs: short for records you may need to change quickly, longer for stable ones.
Treating DNS as a dependency of almost everything — and a common outage cause.
04 · Practice question
“We moved our service to a new IP an hour ago, but some users still reach the old one. Why, and what would you do?”
A strong answer covers
Resolvers cache the old record until its TTL expires; some clients and runtimes also cache by their own rules.
Check the TTL that was published before the change — that’s the minimum wait for well-behaved resolvers.
Keep the old endpoint serving (or forwarding) until its traffic drains.
Next time: lower the TTL well before a planned move, then raise it again afterwards.