Software Engineering Glossary

TTL

Also known as: Time To Live Expiry

Definition

TTL stands for time to live. It is a length of time after which something like a lease, a cache entry, a DNS record, or a token is treated as expired and can be removed or refreshed. TTL is how distributed systems answer the question of how long to trust a piece of data without checking with the source on every read.

Key Takeaways

  • TTL is the difference between fast caching and correct caching. Too long and you serve stale data. Too short and you hammer the source.
  • In coordination, TTL is the safety net that turns a plain lock into a lease that cannot get stuck forever.
  • DNS TTLs decide how fast DNS changes spread across resolvers. Short TTLs mean quick updates but more traffic to the authoritative servers.
  • Always measure TTL against a monotonic clock. Wall clocks can jump back and cause expiry bugs.

How It Works

  1. When a record is created or refreshed, it gets a stamp of expiration time equal to now plus the TTL.
  2. On a read, the system compares now to the expiration. Anything past it is treated as missing or stale.
  3. For leases and locks, the TTL is renewed by heartbeats before it runs out.
  4. Expired records are cleaned up by a background sweeper, on the next read, or by the consistent core itself, like etcd or ZooKeeper.

Where It Is Used

  • DNS A records carry a TTL that controls how long resolvers cache them.
  • Redis EXPIRE and SETEX commands set a TTL for cache invalidation.
  • etcd lease grants take a TTL in seconds and auto delete keys when they are not renewed.
  • Kubernetes lease objects use leaseDurationSeconds and renewTime to compute the effective expiry.