TTL

networking

Stands for: Time To Live

A counter that limits how long data lives or how far a packet may travel before being discarded.

In DNS, TTL is how many seconds a record may be cached before it must be looked up again; in IP, it is the hop limit that stops a lost packet from circling forever. Same name, two jobs, both about bounding lifetime.

Time to live is the same idea implemented in two unrelated places, and conflating them causes real confusion. In an IP packet, it is a hop counter decremented by every router, and when it reaches zero the packet is discarded and an ICMP message is returned. It exists to stop packets circulating forever when routing loops form, and traceroute works entirely by exploiting it, sending packets with deliberately small values to make each hop announce itself.

In DNS, the same acronym means how long a resolver may cache an answer before asking again. This is the one that determines how quickly a change takes effect, and the trade is direct: low values make changes propagate fast and multiply query load, high values are efficient and mean a mistake stays visible for hours.

The operational discipline is lowering DNS values before a planned change, not during the incident it causes. Once a record is cached across the internet at a day-long value, nothing you do makes those resolvers ask again sooner. Lower it well ahead, make the change, then raise it back once the new answer has settled.

Also known as: ttl, time to live, hop limit

← All glossary entries