Cache Expirations

Every cached key can have a time-to-live (TTL): a timer after which the cache deletes it automatically. TTLs keep data fresh and stop memory from filling up forever.

How to use it

await cache.set(
  "popular_posts",
  JSON.stringify(posts),
  { EX: 60 * 5 } // expires in 5 minutes
);

await cache.set("otp:user-123", "482910", { EX: 60 }); // gone after a minute

A TTL is a countdown on a key in memory. Reads hit until it runs out, then the cache deletes the key on its own and the next GET misses.

Freshness vs. staleness

A cached value is a copy, so it can go stale when the real data changes. Two ways to handle it:

Short TTL — accept that data may be a little out of date, and let it refresh on its own
Invalidate on write — delete the cache key whenever the underlying data changes

Even without a TTL, a cache has limited memory. When it's full, it evicts keys to make room (commonly the least recently used). So treat every cached value as something that can disappear at any time.

Don't confuse the three "expirations"

By now you've seen three different things that "expire," and they're easy to mix up. The TTL here is the one that lives in memory:

Expiration	What goes away	Used for
Cache TTL	A value in memory	Freshness
Object expiration	The file itself, after N days	Cost, compliance, temp files
Signed URL expiry	The link; the file stays	Time-limited access

Exercise

Redis challenge · runs in your browser

Key expiration — TTL, EXPIRE, PERSIST

Implement three cache-management functions: • cacheWithExpiry(redis, key, value, ttlSeconds) — store a value that auto-deletes after ttlSeconds • getRemainingTtl(redis, key) — seconds until expiry (-1 = permanent, -2 = does not exist) • makePermanent(redis, key) — remove the expiry so the key never auto-deletes Key expiration is one of Redis's most powerful features. Use it for sessions, rate-limit windows, short-lived tokens, and any data that should auto-clean.

Commands run against an in-browser Redis mock that implements the full ioredis API. Your code works identically against a real Redis instance — just swap in new Redis() from ioredis.