Questions about maxmemory, PERSIST, EXPIRE, Eviction Policy, Expirations

[wernermorgenstern]

So, I have a discussion with a coworker about how DragonFly is handling Evictions and expirations.

I want to get clarification on how DragonFly handles those.
We are using DragonFly in AWS EKS.

We have Cache Mode enabled.
We have some keys which we want to expire, and we use the EXPIRE command to set the expiration.
Some keys we never want to expire, and for those, we execute the PERSIST command.

Now for the two different scenarios.

1. maxmemory=0 is used.
   This means that DragonFly will see how much memory is available, and allocate that memory.

2. maxmemory=2gb (for example, 2gb)
   This means that we set the Maximum Memory explicitly to a value.

So here is my question.
If we reach Maximum Memory in the keys being used (OOM), how are keys expired and evicted?
If maxmemory=0, will keys still be evicted, even if the keys are not expired yet?
Or will it only be done, when maxmemory=VALUE is being used?

We want to know, if we want to have a similar behaviour what Redis has for volatile-ttl, how do we need to configure DragonFly?

volatile-ttl: Evict keys with the expire field set to true that have the shortest remaining time-to-live (TTL) value.

The way Redis does it, even if there are keys, where the expiration has not been reached yet, but we are reaching an OOM situation, it will evict keys which have the shortest/closed remaining TTL value.

The discussion I have with my coworker, is that when we use maxmemory=0, then the evictions will not take place.

[wernermorgenstern]

@romange , not sure if you can help with that please.