Mastering Redis Cache Expiration: Commands, Lazy Deletion & Scheduled Cleanup

1 Introduction

Through previous chapters we know Redis is a KV database stored in memory, but memory is limited. To clean unnecessary data we can delete directly or set expiration. Expiration does not delete immediately; Redis removes expired data via two methods.

Lazy deletion

Periodic deletion via scheduled tasks

2 Redis Expiration Commands

We use commands to set expiration for a key; otherwise the key persists until explicitly deleted.

<code># Cache expiration command example
EXPIRE key seconds [NX|XX|GT|LT]
</code>

Since Redis 7.0, EXPIRE supports options:

NX: set expiration only if the key has no expiration

XX: set expiration only if the key already has expiration

GT: set new expiration only if it is greater than the current one

LT: set new expiration only if it is less than the current one

GT, LT and NX are mutually exclusive. Official test case:

<code>redis> SET mykey "Hello"
"OK"
redis> EXPIRE mykey 10
(integer) 1
redis> TTL mykey
(integer) 10
redis> SET mykey "Hello World"
"OK"
redis> TTL mykey
(integer) -1
redis> EXPIRE mykey 10 XX
(integer) 0
redis> TTL mykey
(integer) -1
redis> EXPIRE mykey 10 NX
(integer) 1
redis> TTL mykey
(integer) 10
</code>

3 Two Deletion Methods for Expired Data

Redis deletes expired data via:

Lazy deletion

Periodic deletion

3.1 Lazy Deletion

When a client accesses a key, Redis checks expiration; if expired, it deletes the key and fetches fresh data. This puts deletion responsibility on client requests, so keys that are never accessed may remain expired for a long time.

The source function expireIfNeeded implements this logic:

<code>int expireIfNeeded(redisDb *db, robj *key, int force_delete_expired) {
    // Return 0 for unexpired keys
    if (!keyIsExpired(db,key)) return 0;
    /* Slave handling omitted for brevity */
    if (server.masterhost != NULL) {
        if (server.current_client == server.master) return 0;
        if (!force_delete_expired) return 1;
    }
    if (checkClientPauseTimeoutAndReturnIfPaused()) return 1;
    deleteExpiredKeyAndPropagate(db,key);
    return 1;
}
</code>

3.2 Periodic Deletion

Relying solely on lazy deletion is insufficient because some keys may never be accessed. Redis also runs a periodic task, frequency configured by the hz setting in redis.conf (default 10 per second).

<code># Run 10 times per second
hz 10
</code>

The task randomly samples a subset of keys with expiration, checks them, and deletes those that are expired. The process:

serverCron executes based on hz .

It scans only keys that have an expiration time.

To avoid long pauses, it scans a limited number of keys per bucket (default 20).

Scanning continues across hash buckets until the limit is reached.

Expired keys found are deleted.

If more than 25% of scanned keys are expired, the cycle repeats.

Additional notes:

Scanning all keys would be costly; Redis aims to stay fast.

In cluster mode, both lazy and periodic deletions are recorded in AOF and propagated to slaves.

4 Summary

Both lazy and periodic deletions may leave some expired keys undeleted, especially if they exceed 25% of scanned keys and are never accessed, potentially exhausting memory. Redis therefore provides a comprehensive memory eviction policy, which will be covered in the next section.