TTL และการหมดอายุใน Redis

Cached data goes stale. Time To Live (TTL) lets Redis automatically remove a key after a set duration, so your cache stays reasonably fresh without manual cleanup.

The EXPIRE command sets a timeout in seconds on an existing key. After that time, Redis deletes the key automatically.

You can set a value and its TTL together. SET key value EX seconds (or the older SETEX) avoids a race between writing and expiring.

The TTL command returns seconds left. PTTL returns milliseconds. Special return values matter:

• -1: key exists but has no expiry
• -2: key does not exist

The PERSIST command removes the TTL from a key, making it permanent again. Useful when a value should stop expiring after some condition is met.

Redis does not check every key constantly. With lazy expiry, an expired key is only removed when someone tries to access it. Until then it lingers in memory.

To reclaim memory from keys nobody touches, Redis also runs a background sampler that periodically removes a batch of expired keys. This is active expiry. Together with lazy expiry it keeps memory in check.

TTL is voluntary removal at a set time. Eviction happens when Redis hits its maxmemory limit and must drop keys to make room. They are different mechanisms that often work together.

Eviction policies decide what to drop under memory pressure:

• volatile-lru: evict least-recently-used keys that have a TTL
• allkeys-lru: evict LRU among all keys

For a pure cache, allkeys policies are common.

Pick TTLs from the data's tolerance for staleness: seconds for fast-changing prices, hours for catalogs, days for rarely-changing config. Add jitter to avoid synchronized mass expiry.

Forgetting to set a TTL turns a cache into an ever-growing store that eventually exhausts memory. For cache use cases, almost every key should have an expiry or rely on an allkeys eviction policy.

You learned TTL and expiration in Redis:

• EXPIRE / SET ... EX set timeouts; TTL/PTTL inspect them; PERSIST removes them.
• Redis uses lazy plus active expiration.
• Eviction policies (volatile vs allkeys) handle memory pressure.
• Always set TTLs (or an allkeys policy) for cache keys.

Good expiration keeps a cache both fresh and bounded.