Updated Mar 23, 2026

Created Mar 23, 2026

Changing the Redis eviction policy

You can change the Redis maxmemory-policy with hypernode-systemctl settings redis_eviction_policy. This policy determines which keys Redis can evict when memory is full.

Available policies

The following values are supported for redis_eviction_policy:

• noeviction

• allkeys-lru

• allkeys-lfu

• allkeys-random

• volatile-lru

• volatile-lfu

• volatile-ttl

• volatile-random

Check the current policy

To view the current Redis eviction policy:

hypernode-systemctl settings redis_eviction_policy

Change the policy

Use the following command to change the Redis eviction policy:

hypernode-systemctl settings redis_eviction_policy <policy>

For example:

hypernode-systemctl settings redis_eviction_policy allkeys-lru

Policy explanation

noeviction

Redis will not evict any keys when memory is full. New write operations that require memory will fail.

allkeys-lru

Redis can evict any key and removes the least recently used keys first.

allkeys-lfu

Redis can evict any key and removes the least frequently used keys first.

allkeys-random

Redis can evict any key and removes keys at random.

volatile-lru

Redis only evicts keys with an expiration time and removes the least recently used keys first.

This is the default policy on Hypernode.

volatile-lfu

Redis only evicts keys with an expiration time and removes the least frequently used keys first.

volatile-ttl

Redis only evicts keys with an expiration time and prefers keys that will expire soonest.

volatile-random

Redis only evicts keys with an expiration time and removes those keys at random.

Important considerations

• Eviction only takes place when Redis reaches its memory limit.

• Policies starting with allkeys- can evict any key, including keys without a TTL.

• Even with an allkeys- policy, Redis can still run into memory pressure if keys cannot be evicted fast enough to keep up with writes.

• Setting a TTL on cache keys is still recommended, even when using an allkeys- policy.

• Policies starting with volatile- only apply to keys with a TTL.

• If not enough keys have a TTL, Redis may not be able to evict enough data with a volatile- policy, and write operations may fail.

• Changing the policy does not remove existing keys immediately.